Prospective validation of dermoscopy-based open-source artificial intelligence for melanoma diagnosis (PROVE-AI study).

The use of artificial intelligence (AI) has the potential to improve the assessment of lesions suspicious of melanoma, but few clinical studies have been conducted. We validated the accuracy of an open-source, non-commercial AI algorithm for melanoma diagnosis and assessed its potential impact on dermatologist decision-making. We conducted a prospective, observational clinical study to assess the diagnostic accuracy of the AI algorithm (ADAE) in predicting melanoma from dermoscopy skin lesion images. The primary aim was to assess the reliability of ADAE's sensitivity at a predefined threshold of 95%. Patients who had consented for a skin biopsy to exclude melanoma were eligible. Dermatologists also estimated the probability of melanoma and indicated management choices before and after real-time exposure to ADAE scores. All lesions underwent biopsy. Four hundred thirty-five participants were enrolled and contributed 603 lesions (95 melanomas). Participants had a mean age of 59 years, 54% were female, and 96% were White individuals. At the predetermined 95% sensitivity threshold, ADAE had a sensitivity of 96.8% (95% CI: 91.1-98.9%) and specificity of 37.4% (95% CI: 33.3-41.7%). The dermatologists' ability to assess melanoma risk significantly improved after ADAE exposure (AUC 0.7798 vs. 0.8161, p = 0.042). Post-ADAE dermatologist decisions also had equivalent or higher net benefit compared to biopsying all lesions. We validated the accuracy of an open-source melanoma AI algorithm and showed its theoretical potential for improving dermatology experts' ability to evaluate lesions suspicious of melanoma. Larger randomized trials are needed to fully evaluate the potential of adopting this AI algorithm into clinical workflows.

Management of immune-related cutaneous adverse events with dupilumab.

Immune checkpoint inhibitors (ICI) target the PD-1/PD-L1 and CTLA-4 pathways and allows the immune system to deliver antitumor effects. However, it is also associated with well-documented immune-related cutaneous adverse events (ircAEs), affecting up to 70-90% of patients on ICI. In this study, we describe the characteristics of and patient outcomes with ICI-associated steroid-refractory or steroid-dependent ircAEs treated with dupilumab. Patients with ircAEs treated with dupilumab between March 28, 2017, and October 1, 2021, at Memorial Sloan Kettering Cancer Center were included in this retrospective study, which assessed the rate of clinical response of the ircAE to dupilumab and any associated adverse events (AEs). Laboratory values were compared before and after dupilumab. All available biopsies of the ircAEs were reviewed by a dermatopathologist. Thirty-four of 39 patients (87%, 95% CI: 73% to 96%) responded to dupilumab. Among these 34 responders, 15 (44.1%) were complete responders with total ircAE resolution and 19 (55.9%) were partial responders with significant clinical improvement or reduction in severity. Only 1 patient (2.6%) discontinued therapy due to AEs, specifically, injection site reaction. Average eosinophil counts decreased by 0.2 K/mcL (p=0.0086). Relative eosinophils decreased by a mean of 2.6% (p=0.0152). Total serum immunoglobulin E levels decreased by an average of 372.1 kU/L (p=0.0728). The most common primary inflammatory patterns identified on histopathological examination were spongiotic dermatitis (n=13, 33.3%) and interface dermatitis (n=5, 12.8%). Dupilumab is a promising option for steroid-refractory or steroid-dependent immune-related cutaneous adverse events, particularly those that are eczematous, maculopapular, or pruritic. Among this cohort, dupilumab was well-tolerated with a high overall response rate. Nonetheless, prospective, randomized, controlled trials are warranted to confirm these observations and confirm its long-term safety.

Agreement Between Experts and an Untrained Crowd for Identifying Dermoscopic Features Using a Gamified App: Reader Feasibility Study.

BACKGROUND: Dermoscopy is commonly used for the evaluation of pigmented lesions, but agreement between experts for identification of dermoscopic structures is known to be relatively poor. Expert labeling of medical data is a bottleneck in the development of machine learning (ML) tools, and crowdsourcing has been demonstrated as a cost- and time-efficient method for the annotation of medical images. OBJECTIVE: The aim of this study is to demonstrate that crowdsourcing can be used to label basic dermoscopic structures from images of pigmented lesions with similar reliability to a group of experts. METHODS: First, we obtained labels of 248 images of melanocytic lesions with 31 dermoscopic "subfeatures" labeled by 20 dermoscopy experts. These were then collapsed into 6 dermoscopic "superfeatures" based on structural similarity, due to low interrater reliability (IRR): dots, globules, lines, network structures, regression structures, and vessels. These images were then used as the gold standard for the crowd study. The commercial platform DiagnosUs was used to obtain annotations from a nonexpert crowd for the presence or absence of the 6 superfeatures in each of the 248 images. We replicated this methodology with a group of 7 dermatologists to allow direct comparison with the nonexpert crowd. The Cohen κ value was used to measure agreement across raters. RESULTS: In total, we obtained 139,731 ratings of the 6 dermoscopic superfeatures from the crowd. There was relatively lower agreement for the identification of dots and globules (the median κ values were 0.526 and 0.395, respectively), whereas network structures and vessels showed the highest agreement (the median κ values were 0.581 and 0.798, respectively). This pattern was also seen among the expert raters, who had median κ values of 0.483 and 0.517 for dots and globules, respectively, and 0.758 and 0.790 for network structures and vessels. The median κ values between nonexperts and thresholded average-expert readers were 0.709 for dots, 0.719 for globules, 0.714 for lines, 0.838 for network structures, 0.818 for regression structures, and 0.728 for vessels. CONCLUSIONS: This study confirmed that IRR for different dermoscopic features varied among a group of experts; a similar pattern was observed in a nonexpert crowd. There was good or excellent agreement for each of the 6 superfeatures between the crowd and the experts, highlighting the similar reliability of the crowd for labeling dermoscopic images. This confirms the feasibility and dependability of using crowdsourcing as a scalable solution to annotate large sets of dermoscopic images, with several potential clinical and educational applications, including the development of novel, explainable ML tools.

Autoimmune Bullous Skin Disorders with Immune Checkpoint Inhibitors Targeting PD-1 and PD-L1.

Monoclonal antibodies (mAb) targeting immune checkpoint pathways such as cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and programmed death 1 (PD-1) may confer durable disease control in several malignancies. In some patients, immune checkpoint mAbs cause cutaneous immune-related adverse events. Although the most commonly reported cutaneous toxicities are mild, a subset may persist despite therapy and can lead to severe or life-threatening toxicity. Autoimmune blistering disorders are not commonly associated with immune checkpoint mAb therapy. We report a case series of patients who developed bullous pemphigoid (BP), an autoimmune process classically attributed to pathologic autoantibody formation and complement deposition. Three patients were identified. Two patients developed BP while receiving the anti-PD-1 mAb nivolumab, and one while receiving the anti-PD-L1 mAb durvalumab. The clinicopathologic features of each patient and rash, and corresponding radiologic findings at the development of the rash and after its treatment, are described. Patients receiving an anti-PD-1/PD-L1 mAb may develop immune-related BP. This may be related to both T-cell- and B-cell-mediated responses. Referral to a dermatologist for accurate diagnosis and management is recommended. Cancer Immunol Res; 4(5); 383-9. ©2016 AACR.

Technology and Technique Standards for Camera-Acquired Digital Dermatologic Images: A Systematic Review.

IMPORTANCE: Photographs are invaluable dermatologic diagnostic, management, research, teaching, and documentation tools. Digital Imaging and Communications in Medicine (DICOM) standards exist for many types of digital medical images, but there are no DICOM standards for camera-acquired dermatologic images to date. OBJECTIVE: To identify and describe existing or proposed technology and technique standards for camera-acquired dermatologic images in the scientific literature. EVIDENCE REVIEW: Systematic searches of the PubMed, EMBASE, and Cochrane databases were performed in January 2013 using photography and digital imaging, standardization, and medical specialty and medical illustration search terms and augmented by a gray literature search of 14 websites using Google. Two reviewers independently screened titles of 7371 unique publications, followed by 3 sequential full-text reviews, leading to the selection of 49 publications with the most recent (1985-2013) or detailed description of technology or technique standards related to the acquisition or use of images of skin disease (or related conditions). FINDINGS: No universally accepted existing technology or technique standards for camera-based digital images in dermatology were identified. Recommendations are summarized for technology imaging standards, including spatial resolution, color resolution, reproduction (magnification) ratios, postacquisition image processing, color calibration, compression, output, archiving and storage, and security during storage and transmission. Recommendations are also summarized for technique imaging standards, including environmental conditions (lighting, background, and camera position), patient pose and standard view sets, and patient consent, privacy, and confidentiality. Proposed standards for specific-use cases in total body photography, teledermatology, and dermoscopy are described. CONCLUSIONS AND RELEVANCE: The literature is replete with descriptions of obtaining photographs of skin disease, but universal imaging standards have not been developed, validated, and adopted to date. Dermatologic imaging is evolving without defined standards for camera-acquired images, leading to variable image quality and limited exchangeability. The development and adoption of universal technology and technique standards may first emerge in scenarios when image use is most associated with a defined clinical benefit.

Effects on skills and practice from a web-based skin cancer course for primary care providers.

BACKGROUND: Melanoma incidence and mortality is a growing concern. Better recognition and management of skin cancer by primary care providers (PCPs) could help, but studies suggest they would benefit from additional education. Effective educational programs are needed. METHODS: We developed and conducted a voluntary before-and-after evaluation of a 1- to 2-hour interactive, web-based course in skin cancer detection for practicing, board-certified PCPs (http://www.skinsight.com/info/for_professionals/dermatology-education-resources). Voluntary participants' ability to diagnose and manage skin cancer was assessed using pretests, immediate tests, and 6-month posttests. The effect on actual practice patterns was assessed using participants' patient panels: referrals or visits to dermatology and skin biopsies during the 6 months after the course were compared with those during the same period before the course. RESULTS: The mean age of the 54 participants was 50.5 years (standard deviation, 11.1); 54% were women and 52% were Asian. The mean score for appropriate diagnosis and management increased from 36.1% to 46.7% (odds ratio, 1.6; 95% confidence interval, 1.4-1.9), with greatest improvement in benign lesions, from 32.1% to 46.3% (odds ratio, 1.9; 95% confidence interval, 1.6-2.4). Dermatology referrals for suspicious lesions or new visits by participants' patients decreased at both sites after the course (from 630 to 607 and from 726 to 266, respectively). CONCLUSIONS: This course improved skills in practicing PCPs. Improvement was greatest in the diagnosis and appropriate management of benign lesions and dermatology utilization decreased.

Microinvasive melanoma: cutaneous pharmacotherapeutic approaches.

Surgical excision is the treatment of choice for primary melanomas and radiation therapy is the accepted alternative for the subset of lesions not amenable to surgery. With the recent rise in melanoma incidence, especially in the elderly, there are a growing number of cases that are neither amenable to surgery nor radiation therapy. In this article, we review pharmacotherapeutic approaches to microinvasive melanoma (invasive radial growth phase melanoma) that might be considered in such circumstances. There are no approved drugs for the treatment of primary melanoma and randomized controlled trials with 5 or more years of follow-up have not been performed. The limited studies and numerous case series in the literature on pharmacologic treatment of primary melanoma have focused on topical therapies. Accordingly, we provide a review of the potential pharmacotherapeutic agents in the treatment of microinvasive melanoma by extrapolating from the available limited literature on the use of fluorouracil, azelaic acid, retinoic acid derivatives, interferon (IFN)-α, imiquimod, and other agents for melanoma in situ, invasive melanoma, and epidermotropic melanoma metastases. Our review indicates that topical fluorouracil and tretinoin are not effective as single agents. The efficacy of azelaic acid, tazarotene, cidofovir, and intralesional IFN-α, interleukin-2, and IFN-ß is undefined. Imiquimod is the most studied and promising agent; however, optimal dosage, therapeutic regimen, and survival rates are unknown. In the face of a growing demand for non-surgical treatments, formal clinical trials are needed to ascertain the role of pharmacotherapeutic agents in the treatment of microinvasive melanoma.

Influence of time on dermoscopic diagnosis and management.

BACKGROUND/OBJECTIVES: Dermoscopy aids in clinical decision-making. However, time pressure is a common reason precluding its use. We evaluated the effect of time on lesion recognition and management decisions utilising clinical and dermoscopic images. METHOD: In all, 100 dermoscopic images were presented to 15 dermatologists with experience in dermoscopy and seven non-experts (dermatology residents). Each lesion was displayed thrice in succession. The dermoscopic image was initially presented for 1 s (t1). The same dermoscopic image was shown again without time constraints (t2) and then a final time with additional images of the clinical context (t3). Participants provided a diagnosis, their level of confidence and biopsy predilection after evaluating each image. RESULTS: For benign lesions, both groups rarely changed their diagnosis. However, an improvement in the number of correct benign diagnoses was observed when the lesion was shown in a clinical context. For malignant lesions, both groups improved when more time and clinical context was given; nevertheless, non-experts were more likely to change the diagnosis towards the correct one as more time was given and tended to perform more biopsies, in particular of benign lesions. Limitations were a small number of participants and an artificial study setting. CONCLUSION: Dermoscopy uses analytical and non-analytical reasoning approaches. We suggest that non-analytical reasoning is employed when rapid clinical decisions need to be made, especially during the evaluation of benign lesions. We conclude that dermoscopy is relatively rapid and non-time-consuming technique that adds relevant information and guides clinicians towards appropriate management decisions.

Developing an interactive web-based learning program on skin cancer: the learning experiences of clinical educators.

Web-based learning in medical education is rapidly growing. However, there are few firsthand accounts on the rationale for and development of web-based learning programs. We present the experience of clinical educators who developed an interactive online skin cancer detection and management course in a time-efficient and cost-efficient manner without any prior skills in computer programming or technical construction of web-based learning programs. We review the current state of web-based learning including its general advantages and disadvantages as well as its specific utility in dermatology. We then detail our experience in developing an interactive online skin cancer curriculum for primary care clinicians. Finally, we describe the main challenges faced and lessons learned during the process. This report may serve medical educators who possess minimal computer programming and web design skills but want to employ the many strengths of web-based learning without the huge costs associated with hiring a professional development team.

Skin cancer education for primary care physicians: a systematic review of published evaluated interventions.

BACKGROUND: Early detection of melanoma may provide an opportunity to positively impact melanoma mortality. Numerous skin cancer educational interventions have been developed for primary care physicians (PCPs) to improve diagnostic accuracy. Standardized training is also a prerequisite for formal testing of melanoma screening in the primary care setting. OBJECTIVE: We conducted a systematic review to determine the extent of evaluated interventions designed to educate PCPs about skin cancer, including melanoma. DESIGN: Relevant studies in the English language were identified through systemic searches performed in MEDLINE, EMBASE, BIOSIS, and Cochrane through December 2010. Supplementary information was obtained from corresponding authors of the included studies when necessary. APPROACH: Studies eligible for inclusion formally evaluated skin cancer education interventions and were designed primarily for PCPs. Excluded studies lacked a specified training intervention, used decision-making software, focused solely on risk factor identification, or did not directly educate or assess participants. Twenty studies met the selection criteria. Data were extracted according to intervention content and delivery format, and study outcomes. KEY RESULTS: All interventions included instructions about skin cancer diagnosis, but otherwise varied in content. Curricula utilized six distinct educational techniques, usually incorporating more than one. Intervention duration varied from 12 min to over 6 h. Eight of the 20 studies were randomized trials. Most studies (18/20, 90%) found a significant improvement in at least one of the following five outcome categories: knowledge, competence, confidence, diagnostic performance, or systems outcomes. Competence was most commonly measured; no study evaluated all categories. Variability in study design, interventions, and outcome measures prevented correlation of outcomes with intervention characteristics. CONCLUSIONS: Despite the development of many isolated educational interventions, few have been tested rigorously or evaluated under sufficient standardized conditions to allow for quantitative comparison. Improved and rigorously tested skin cancer educational interventions for PCPs with outcome measures focusing on changes in performance are needed.