BT33 Ethical challenges with skin cancer and artificial intelligence integration

Sut Mo Zachary Chan,Faishal Dubash

DOI: https://doi.org/10.1093/bjd/ljae090.430

IF: 11.113

2024-06-28

British Journal of Dermatology

Abstract:Abstract The integration of artificial intelligence (AI) in the field of dermatology promises a revolutionary transformation in our approach to this domain. However, the use of AI in dermatology is still in its early stages, and with limited scope of applications. While some of the existing research performed has highlighted the capabilities of AI in diagnosing lesions comparable with those of a dermatologist, the implementation of AI brings about its own ethical challenges that affect both the clinician and the patient. This article aims to explore the ethical issues and prospects surrounding AI implementation in dermatology by drawing insights from the current literature. A qualitative analysis was conducted on articles focusing on AI usage in medical care, with a specific emphasis on dermatology. The analysis unveils several ethical issues that emerge from the implementation of AI in dermatology: the obscure and subjective nature of AI leads to selective bias and unequal treatment based on skin tone differences. It also has the potential to erode patient autonomy by taking on a more paternalistic approach to medicine, increasing the risk of unwarranted harm to patients because of unnecessary biopsies, along with potential misalignment of treatment priorities between AI systems and patients. Issues of accountability arise as AI integration can challenge the epistemic authority and ownership of clinicians over their patients. Finally, the closed-loop system of training algorithms coupled with the lack of an open-access database due to privacy issues further increases the subjectiveness and inaccessibility of each algorithm. While AI’s integration in dermatology promises to streamline the diagnostic process, it also sparks ethical issues that affect the main stakeholders. These findings indicate the need for further development and enhancement of AI systems before they are capable of being implemented. Introduction of a transparent, explainable AI model would alleviate concerns regarding patient autonomy and treatment rationalization. Simultaneously, the establishment of a global open-access database will serve to mitigate selective bias. Furthermore, additional research comparing algorithms would be useful in establishing a standardized validation tool. As it stands, rather than replacing clinicians, AI should be used as a diagnostic adjunct in aiding clinical decisions.

dermatology