Photo Credit: Ismagilov
AI use is expanding in dermatology, but experts caution that more research is needed to optimize AI’s functionality for patients with skin of color.
AI is changing how clinicians care for their patients, but problems with the technology persist. One major area of interest is AI in dermatology care; however, experts say AI offers weak support when caring for patients with skin of color.
Physician’s Weekly (PW) spoke with lead author Rebecca Fliorent about her team’s literature review regarding updates and challenges related to AI and dermatological care for patients with skin of color. The review was published in the International Journal of Dermatology.
PW: Why was it important to research this topic?
Fliorent: After conducting studies on racial disparities in dermatology, it became evident that further research is crucial. With the growing prevalence of AI in medicine, particularly in dermatology, it is essential to evaluate how effectively AI tools detect skin lesions in patients with skin of color. This research is vital because accurate detection and diagnosis in diverse populations can help reduce health disparities and ensure that AI advancements benefit all patients equitably.
What were your materials, methods, and population?
AI programs in dermatology are being developed at a rapid pace. In our study, we examined the dermatology AI programs available at the time and evaluated whether their efficacy in diagnosing lesions had been thoroughly studied, with a particular focus on whether their data included patients with skin of color.
Additionally, we examined various factors contributing to the challenges of applying AI in dermatology for skin of color, including the skin tone scales used by AI algorithms and the quality and standardization of images these algorithms rely on.
What are some highlights from your data, and what is your interpretation of these results?
Our study identified significant challenges in applying AI to skin of color, particularly concerning evaluating current AI algorithms, reliance on the Fitzpatrick skin tone scale (FST), and image quality and standardization issues.
The FST scale does not represent the full spectrum of skin tones, leading to biased data and potentially inaccurate health assessments. The Monk scale, which is gaining traction in the industry with support from companies like Google, offers a more inclusive approach to categorizing skin tones. Additionally, factors such as inconsistent lighting and improper camera settings when photographing skin of color contribute to poor image quality, further diminishing the accuracy of AI in dermatology.
To ensure AI in dermatology is both effective and equitable, it is crucial to adopt more inclusive skin tone scales like the Monk scale, enhance skin of color representation in training datasets, and standardize medical photography practices.
What are the most important things for clinicians to know about your work?
As medicine advances, so does technology. While AI programs are not perfect, we have the potential to refine these tools to better assist dermatologists in diagnosing suspicious lesions. Additionally, we hope that enhancing AI algorithms will empower patients by educating them on recognizing potential skin lesions early, ultimately leading to more timely interventions and improved health outcomes.
What questions remain? Do you have a follow-up study coming?
Many AI algorithms lack clinical data. It is essential to focus future research on evaluating these technologies from a patient-centered and clinical perspective. AI algorithms should also aim to incorporate more diverse images of patients with skin of color.
My goal is to collaborate with software developers and leaders in AI in dermatology to enhance the quality of patient care and help reduce racial disparities.
Is there anything else you would like to add?
Patients with skin of color have historically been underrepresented in medical literature, making it crucial to ensure their inclusion in our AI algorithms and research. It is also important to emphasize that these AI programs aim not to replace dermatologists but to enhance workflow efficiency and improve patient health literacy.
