Progress and challenges of artificial intelligence in skin of color

Artificial intelligence (AI) has marked a paradigm shift in dermatology by enabling computers to simulate human-like problem-solving and decision-making. AI algorithms and models learn from data, recognize patterns, and make informed decisions. This helps to identify features and patterns of various dermatologic conditions, including aiding in the early detection of skin cancer. AI can also personalize treatment recommendations based on patient data, including medical history, symptoms, and treatment outcomes. This technology is a valuable adjunct to dermatologists, enhancing diagnostic accuracy and clinical decision-making, which can significantly improve patient outcomes.1 Despite remarkable progress, challenges still persist, particularly in applying AI to individuals with Skin of Color (SOC). A recent review by Fliorent et al.2 stated that the challenges associated with applying AI to SOC in the context of dermatology arise from several factors, including the constrained scope of the Fitzpatrick skin type (FST), insufficient representation of SOC in datasets, and concerns related to image quality and standardization. Due to these existing issues, current AI programs face difficulties in effectively identifying lesions in SOC. Moreover, it was observed that around 30% of the identified programs lacked reported data on their utilization in dermatology, particularly in the context of SOC. Dermatologic research has historically neglected populations with SOC, resulting in limited data and studies addressing their unique clinical presentations. The inadequate attention given to dermatologic conditions specific to individuals with SOC in medical education magnifies the absence of awareness and evidence-based data for these particular populations.3 To enhance AI's effectiveness in dermatology for SOC, several key steps are necessary. First, the reliance on the FST for skin type classification should be reevaluated. The Monk scale, designed to capture variations in human skin pigmentation, offers a more inclusive categorization. Collaborations with entities like Google and Facebook/Meta, which incorporate the Monk scale, can contribute to a more diverse representation in AI algorithms.4 Addressing the underrepresentation of SOC in datasets requires intentional efforts to diversify training data. Training AI models on artificially "darkened" images is one approach, but incorporating authentic images of SOC lesions directly into datasets is crucial for accuracy. Moreover, image quality issues, including consistent lighting and standardized photography techniques, must be addressed to ensure the reliability of AI algorithms across diverse skin tones. A comprehensive review of current AI programs in dermatology reveals both promise and limitations. Programs like VisualDx and DermAssist aim to enhance diagnostic accuracy, but their effectiveness in SOC patients requires further development. The need for diverse datasets is evident in the varied success rates of AI algorithms like ModelDerm, DeepDerm, and HAM 10000, emphasizing the urgency of addressing representation issues.5 As we look to the future, AI has enormous potential for reducing health gaps in dermatology. Future applications include integrating AI into primary care settings, emergency rooms, and urgent care centers to triage dermatologic referrals, particularly in regions lacking dermatologists. This expansion can benefit individuals with limited access to dermatological care, promoting more equitable health outcomes. In conclusion, AI integration into dermatologic practice represents a revolutionary advancement toward precision medicine. While challenges persist, especially in addressing the unique needs of SOC populations, efforts in data diversification, image quality enhancement, and program development can advance AI to new heights in dermatology. By encouraging inclusivity and integrating advancements in technology, the dermatologic community can fully utilize the potential of AI for the well-being of every patient, regardless of their skin color.