Human-AI interaction in skin cancer diagnosis: a systematic review and meta-analysis.

The development of diagnostic tools for skin cancer based on artificial intelligence (AI) is increasing rapidly and will likely soon be widely implemented in clinical use. Even though the performance of these algorithms is promising in theory, there is limited evidence on the impact of AI assistance on human diagnostic decisions. Therefore, the aim of this systematic review and meta-analysis was to study the effect of AI assistance on the accuracy of skin cancer diagnosis. We searched PubMed, Embase, IEE Xplore, Scopus and conference proceedings for articles from 1/1/2017 to 11/8/2022. We included studies comparing the performance of clinicians diagnosing at least one skin cancer with and without deep learning-based AI assistance. Summary estimates of sensitivity and specificity of diagnostic accuracy with versus without AI assistance were computed using a bivariate random effects model. We identified 2983 studies, of which ten were eligible for meta-analysis. For clinicians without AI assistance, pooled sensitivity was 74.8% (95% CI 68.6-80.1) and specificity was 81.5% (95% CI 73.9-87.3). For AI-assisted clinicians, the overall sensitivity was 81.1% (95% CI 74.4-86.5) and specificity was 86.1% (95% CI 79.2-90.9). AI benefitted medical professionals of all experience levels in subgroup analyses, with the largest improvement among non-dermatologists. No publication bias was detected, and sensitivity analysis revealed that the findings were robust. AI in the hands of clinicians has the potential to improve diagnostic accuracy in skin cancer diagnosis. Given that most studies were conducted in experimental settings, we encourage future studies to further investigate these potential benefits in real-life settings.

Telehealth Utilization and Associations in the United States During the Third Year of the COVID-19 Pandemic: Population-Based Survey Study in 2022.

BACKGROUND: The COVID-19 pandemic rapidly changed the landscape of clinical practice in the United States; telehealth became an essential mode of health care delivery, yet many components of telehealth use remain unknown years after the disease's emergence. OBJECTIVE: We aim to comprehensively assess telehealth use and its associated factors in the United States. METHODS: This cross-sectional study used a nationally representative survey (Health Information National Trends Survey) administered to US adults (≥18 years) from March 2022 through November 2022. To assess telehealth adoption, perceptions of telehealth, satisfaction with telehealth, and the telehealth care purpose, we conducted weighted descriptive analyses. To identify the subpopulations with low adoption of telehealth, we developed a weighted multivariable logistic regression model. RESULTS: Among a total of 6252 survey participants, 39.3% (2517/6252) reported telehealth use in the past 12 months (video: 1110/6252, 17.8%; audio: 876/6252, 11.6%). The most prominent reason for not using telehealth was due to telehealth providers failing to offer this option (2200/3529, 63%). The most common reason for respondents not using offered telehealth services was a preference for in-person care (527/578, 84.4%). Primary motivations to use telehealth were providers' recommendations (1716/2517, 72.7%) and convenience (1516/2517, 65.6%), mainly for acute minor illness (600/2397, 29.7%) and chronic condition management (583/2397, 21.4%), yet care purposes differed by age, race/ethnicity, and income. The satisfaction rate was predominately high, with no technical problems (1829/2517, 80.5%), comparable care quality to that of in-person care (1779/2517, 75%), and no privacy concerns (1958/2517, 83.7%). Younger individuals (odd ratios [ORs] 1.48-2.23; 18-64 years vs ≥75 years), women (OR 1.33, 95% CI 1.09-1.61), Hispanic individuals (OR 1.37, 95% CI 1.05-1.80; vs non-Hispanic White), those with more education (OR 1.72, 95% CI 1.03-2.87; at least a college graduate vs less than high school), unemployed individuals (OR 1.25, 95% CI 1.02-1.54), insured individuals (OR 1.83, 95% CI 1.25-2.69), or those with poor general health status (OR 1.66, 95% CI 1.30-2.13) had higher odds of using telehealth. CONCLUSIONS: To our best knowledge, this is among the first studies to examine patient factors around telehealth use, including motivations to use, perceptions of, satisfaction with, and care purpose of telehealth, as well as sociodemographic factors associated with telehealth adoption using a nationally representative survey. The wide array of descriptive findings and identified associations will help providers and health systems understand the factors that drive patients toward or away from telehealth visits as the technology becomes more routinely available across the United States, providing future directions for telehealth use and telehealth research.

Transparent medical image AI via an image-text foundation model grounded in medical literature.

Building trustworthy and transparent image-based medical artificial intelligence (AI) systems requires the ability to interrogate data and models at all stages of the development pipeline, from training models to post-deployment monitoring. Ideally, the data and associated AI systems could be described using terms already familiar to physicians, but this requires medical datasets densely annotated with semantically meaningful concepts. In the present study, we present a foundation model approach, named MONET (medical concept retriever), which learns how to connect medical images with text and densely scores images on concept presence to enable important tasks in medical AI development and deployment such as data auditing, model auditing and model interpretation. Dermatology provides a demanding use case for the versatility of MONET, due to the heterogeneity in diseases, skin tones and imaging modalities. We trained MONET based on 105,550 dermatological images paired with natural language descriptions from a large collection of medical literature. MONET can accurately annotate concepts across dermatology images as verified by board-certified dermatologists, competitively with supervised models built on previously concept-annotated dermatology datasets of clinical images. We demonstrate how MONET enables AI transparency across the entire AI system development pipeline, from building inherently interpretable models to dataset and model auditing, including a case study dissecting the results of an AI clinical trial.

Auditing the inference processes of medical-image classifiers by leveraging generative AI and the expertise of physicians.

The inferences of most machine-learning models powering medical artificial intelligence are difficult to interpret. Here we report a general framework for model auditing that combines insights from medical experts with a highly expressive form of explainable artificial intelligence. Specifically, we leveraged the expertise of dermatologists for the clinical task of differentiating melanomas from melanoma 'lookalikes' on the basis of dermoscopic and clinical images of the skin, and the power of generative models to render 'counterfactual' images to understand the 'reasoning' processes of five medical-image classifiers. By altering image attributes to produce analogous images that elicit a different prediction by the classifiers, and by asking physicians to identify medically meaningful features in the images, the counterfactual images revealed that the classifiers rely both on features used by human dermatologists, such as lesional pigmentation patterns, and on undesirable features, such as background skin texture and colour balance. The framework can be applied to any specialized medical domain to make the powerful inference processes of machine-learning models medically understandable.

Principles, applications, and future of artificial intelligence in dermatology.

This paper provides an overview of artificial-intelligence (AI), as applied to dermatology. We focus our discussion on methodology, AI applications for various skin diseases, limitations, and future opportunities. We review how the current image-based models are being implemented in dermatology across disease subsets, and highlight the challenges facing widespread adoption. Additionally, we discuss how the future of AI in dermatology might evolve and the emerging paradigm of large language, and multi-modal models to emphasize the importance of developing responsible, fair, and equitable models in dermatology.

Assessing Biases in Medical Decisions via Clinician and AI Chatbot Responses to Patient Vignettes.

This cross-sectional study compares clinician and artificial intelligence (AI) chatbot responses to patient vignettes used to identify bias in medical decisions.

Risk of Multiple Primary Cancers in Patients With Merkel Cell Carcinoma: A SEER-Based Analysis.

Importance: The risk of subsequent primary cancers after a diagnosis of cutaneous Merkel cell carcinoma (MCC) is not well established. Objective: To evaluate the risk of subsequent primary cancers after the diagnosis of a first primary cutaneous MCC. Design, Setting, and Participants: This cohort study analyzed data from 17 registries of the Surveillance, Epidemiology, and End Results (SEER) Program from January 1, 2000, to December 31, 2019. In all, 6146 patients diagnosed with a first primary cutaneous MCC were identified. Main Outcomes and Measures: The primary outcome was the relative and absolute risks of subsequent primary cancers after the diagnosis of a first primary MCC, which were calculated using the standardized incidence ratio (SIR; ratio of observed to expected cases of subsequent cancer) and the excess risk (difference between observed and expected cases of subsequent cancer divided by the person-years at risk), respectively. Data were analyzed between January 1, 2000, and December 31, 2019. Results: Of 6146 patients with a first primary MCC diagnosed at a median (IQR) age of 76 (66-83) years, 3713 (60.4%) were men, and the predominant race and ethnicity was non-Hispanic White (5491 individuals [89.3%]). Of these patients, 725 (11.8%) developed subsequent primary cancers, with an SIR of 1.28 (95% CI, 1.19-1.38) and excess risk of 57.25 per 10 000 person-years. For solid tumors after MCC, risk was elevated for cutaneous melanoma (SIR, 2.36 [95% CI, 1.85-2.97]; excess risk, 15.27 per 10 000 person-years) and papillary thyroid carcinoma (SIR, 5.26 [95% CI, 3.25-8.04]; excess risk, 6.16 per 10 000 person-years). For hematologic cancers after MCC, risk was increased for non-Hodgkin lymphoma (SIR, 2.62 [95% CI, 2.04-3.32]; excess risk, 15.48 per 10 000 person-years). Conclusions and Relevance: This cohort study found that patients with MCC had an increased risk of subsequently developing solid and hematologic cancers. This increased risk may be associated with increased surveillance, treatment-related factors, or shared etiologies of the other cancers with MCC. Further studies exploring possible common etiological factors shared between MCC and other primary cancers are warranted.

A Social Media-Based Public Health Campaign to Reduce Indoor Tanning in High-Risk Populations.

Introduction: Indoor tanning beds cause more than 450,000 new skin cancers each year, yet their use remains common, with a global indoor tanning prevalence of 10.4%. Social media provides an opportunity for cost-effective, targeted public health messaging. We sought to direct Instagram users at high risk of indoor tanning to accurate health information about the risks of indoor tanning and to reduce indoor tanning bed use. Methods: We disseminated a public health campaign on Instagram on April 6-27, 2022 with 34 video and still-image advertisements. We had 2 target audiences at high risk of indoor tanning: women aged 18-30 years in Kentucky, Nebraska, Ohio, or Tennessee interested in indoor tanning and men aged 18-45 years in California interested in indoor tanning. To evaluate the impact of the campaign, we tracked online metrics, including website visits, and conducted an interrupted time-series analysis of foot traffic data in our target states for all tanning salons documented on SafeGraph from January 1, 2018 to 3 months after the campaign. Results: Our indoor tanning health information advertisements appeared on Instagram feeds 9.1 million times, reaching 1.06 million individuals. We received 7,004 views of our indoor tanning health information landing page (Average Time on Page of 56 seconds). We did not identify a significant impact on foot traffic data on tanning salons. Conclusions: We show the successful use of social media advertising to direct high-risk groups to online health information about indoor tanning. Future research quantifying tanning visits before and after indoor tanning interventions is needed to guide future public health efforts.

Fostering transparent medical image AI via an image-text foundation model grounded in medical literature.

Building trustworthy and transparent image-based medical AI systems requires the ability to interrogate data and models at all stages of the development pipeline: from training models to post-deployment monitoring. Ideally, the data and associated AI systems could be described using terms already familiar to physicians, but this requires medical datasets densely annotated with semantically meaningful concepts. Here, we present a foundation model approach, named MONET (Medical cONcept rETriever), which learns how to connect medical images with text and generates dense concept annotations to enable tasks in AI transparency from model auditing to model interpretation. Dermatology provides a demanding use case for the versatility of MONET, due to the heterogeneity in diseases, skin tones, and imaging modalities. We trained MONET on the basis of 105,550 dermatological images paired with natural language descriptions from a large collection of medical literature. MONET can accurately annotate concepts across dermatology images as verified by board-certified dermatologists, outperforming supervised models built on previously concept-annotated dermatology datasets. We demonstrate how MONET enables AI transparency across the entire AI development pipeline from dataset auditing to model auditing to building inherently interpretable models.

Dissection of medical AI reasoning processes via physician and generative-AI collaboration.

Despite the proliferation and clinical deployment of artificial intelligence (AI)-based medical software devices, most remain black boxes that are uninterpretable to key stakeholders including patients, physicians, and even the developers of the devices. Here, we present a general model auditing framework that combines insights from medical experts with a highly expressive form of explainable AI that leverages generative models, to understand the reasoning processes of AI devices. We then apply this framework to generate the first thorough, medically interpretable picture of the reasoning processes of machine-learning-based medical image AI. In our synergistic framework, a generative model first renders "counterfactual" medical images, which in essence visually represent the reasoning process of a medical AI device, and then physicians translate these counterfactual images to medically meaningful features. As our use case, we audit five high-profile AI devices in dermatology, an area of particular interest since dermatology AI devices are beginning to achieve deployment globally. We reveal how dermatology AI devices rely both on features used by human dermatologists, such as lesional pigmentation patterns, as well as multiple, previously unreported, potentially undesirable features, such as background skin texture and image color balance. Our study also sets a precedent for the rigorous application of explainable AI to understand AI in any specialized domain and provides a means for practitioners, clinicians, and regulators to uncloak AI's powerful but previously enigmatic reasoning processes in a medically understandable way.

Incidence Trends of Primary Cutaneous T-Cell Lymphoma in the US From 2000 to 2018: A SEER Population Data Analysis.

Using SEER database data, this cohort study analyzed cutaneous T-cell lymphoma incidence by tumor subtype, sex, age, race and ethnicity, socioeconomic status, and geography.

A novel pathogenic RHOA variant in a patient with patterned cutaneous hypopigmentation associated with extracutaneous findings.

RHOA-related neuroectodermal syndrome is characterised by linear skin hypopigmentation along Blaschko's lines associated with alopecia, leukoencephalopathy, facial and limb hypoplasia, and ocular, dental, and acral anomalies. Herein, we report a patient with patterned cutaneous hypopigmentation with a similar phenotype due to a novel postzygotic RHOA variant (c.210G>T; p.Arg70Ser). This illustrates that the complexity of the orchestration of morphogenesis and organogenesis can be affected by different variants in the same gene.

Toxic epidermal necrolysis associated with pembrolizumab.

INTRODUCTION: Stevens-Johnson syndrome and toxic epidermal necrolysis are severe cutaneous drug eruptions characterized by epidermal detachment. Pembrolizumab is a monoclonal antibody that binds to the programmed death-1 receptor, and it has been associated with numerous cutaneous adverse side-effects, including Stevens-Johnson syndrome. CASE REPORT: We describe a 63-year-old male with metastatic lung adenocarcinoma who developed a rapidly progressing maculopapular rash three days after a first dose of pembrolizumab. On day 16, the rash affected more than 80% of body surface area with detachment of large sheets of necrolytic epidermis in 30-40% of body surface area. However, the patient only presented with mild mucosal involvement. Histopathologic examination of a skin biopsy showed a subepidermal blister with overlying prominent full thickness epidermal keratinocytic necrosis and a superficial perivascular infiltrate of lymphocytes. A toxic epidermal necrolysis secondary to pembrolizumab was then diagnosed. Management and outcome: In addition to supportive cares, the patient received corticosteroids and cyclosporine. The patient responded rapidly to the immunosuppressant therapy, and nearly complete re-epithelialization was achieved 24 days after the start of the reaction. DISCUSSION: In our review of the literature, 15 other cases of Stevens-Johnson syndrome/toxic epidermal necrolysis were reported with programmed death-1/programmed cell death ligand-1 inhibitors. To our knowledge, this is the first case of toxic epidermal necrolysis secondary to pembrolizumab published in the literature. The American Society of Clinical Oncology guidelines suggest that cyclosporine, in addition to corticosteroids, be initiated when toxic epidermal necrolysis is suspected. Clinicians should be aware of this rare dermatological emergency with the increasing use of pembrolizumab in oncology.