Ethical considerations for artificial intelligence in dermatology: a scoping review

Abstract

The field of dermatology is experiencing the rapid deployment of artificial intelligence (AI), from mobile applications for skin cancer detection to large language models like ChatGPT that can answer generalist or specialist questions about skin diagnoses. With these new applications, ethical concerns have emerged. In this scoping review, we aim to identify the applications of AI to the field of dermatology and to understand their ethical implications. We utilized a multifaceted search approach, searching PubMed, Medline, Cochrane, and Google Scholar for primary literature according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) Extension for Scoping Reviews. Our advanced query included terms related to dermatology, artificial intelligence, and ethical considerations. Our search yielded a total of 202 papers. After initial screening, 68 studies were included. Thirty-two related to clinical image analysis and raised ethical concerns for misdiagnosis, data security, violations of privacy, and replacement of dermatologist jobs. Seventeen discussed limited skin of color representation in datasets leading to potential misdiagnosis in the general population. Nine articles about teledermatology raised ethical concerns, including the exacerbation of health disparities, lack of standardized regulations, informed consent for AI use, and privacy challenges. Seven addressed inaccuracies of responses of large language models. Seven examined attitudes and trust towards AI, with most patients requesting supplemental assessment by a physician to ensure reliability and accountability. Benefits of artificial intelligence integration into clinical practice include increased patient access, improved clinical decision making, efficiency, and many others. However, safeguards must be implemented to ensure ethical applications of artificial intelligence.

Applications of Large Language Models in Pathology

“…Implementation of LLM tools in healthcare settings are plagued by concerns over cost, computing power, and data privacy. Although GPT4 currently performs better than open source LLMs, there are circumstances where open LLMs are preferable, due to lack of adherence to HIPAA (Health Insurance Portability and Accountability Act) regulations, and other data security concerns [109] involving commercial LLMs. Smaller LLMs, such as the open source FastChat-T5 3B-parameter model can be run locally on a personal computer, while preserving patient privacy [36].…”