The last thing anyone needs at the dermatologist is bad advice, but for BIPOC individuals, this happens—if they can get to one in the first place. In dermatology, BIPOC communities are underrepresented. This results in delays and misdiagnoses; but there may be an answer. Developments in digital health are turning the system around. Digital appointments are raising accessibility, AI programs can give diagnoses, and the community can help by contributing to databases. With new technology, the system can improve.
Disparity in BIPOC Skin Care
African Americans are 4 times as likely to be diagnosed with stage IV melanoma and 1.5 times as likely to die of it as Caucasians, due to delayed diagnosis. Dermatologists are in high demand, and BIPOC individuals face obstacles to care often. This starts with the data and education doctors receive. One study by Mayo Clinic Proceedings (DA; [2]) observed doctors’ practice guidelines. “Of the 54 clinical practice guidelines… 61% did not include any keywords related to BIPOC populations…” Most clinicians fail to target BIPOC skin care.
Having recognized this, education is improving. Medical schools are altering curricula to represent BIPOC skin conditions better. Resources building valuable data, such as the Inclusive Skin Color Project, have been established to increase health equity in research. This data is used for digital treatment options which make dermatology easy. One event has driven improvement more than ever before.
How 2020 Changed Everything
Daniele Giansanti, a well-known researcher in dermatology, studied the effects of COVID-19 on digital health. She quotes: “…during the COVID-19 pandemic, TD [tele-dermatology] integration with mHealth [mobile health] advanced rapidly. AI and mobile apps have empowered citizens to take an active role in their healthcare.” Designed for signs of COVID-19, these apps were applied to dermatology. It’s easy to see why TD apps became a focus of development. With high-performing TD apps, clinicians can broaden their reach and clear non-clinical concerns much more quickly. This allows them to treat those neglected due to lack of access. Digital healthcare is the most promising first step to equal dermatology treatment.
The main goal now is to ensure these digital solutions address BIPOC dermatology, and that they work on a large scale, particularly for low-middle income countries. According to Mayo Clinic Proceedings, “The scarcity of dermatologists in LMICs, with less than 1 per million population, is a significant challenge. Leveraging DHIs is suggested to fill this gap (Skin and Digital–the 2024 Narrative).”
Other challenges and drawbacks should be resolved to benefit the BIPOC population. Widespread use faces issues like internet access and tech literacy. Changes are developing for the best possible patient experience. Each of these problems hinder dermatology AI systems, too. As AI gets faster, it’s important to be aware of its possible downsides.
Challenges for Dermatology AI
Dermatology AI is promising—and complex. AI-driven tools can perform as well as doctors—sometimes. However, the data shows BIPOC underrepresentation. Early tests used minimal data for BIPOC skin conditions(Brinker et. al). In another study, scientists caused racial bias by attempting to avoid it (Aggarwal P). In the AI’s design, they left out racial labels. The AI underestimated BIPOC care needs, because these communities generally spend less on healthcare. Not because they need it less; because they receive it less. Careful tests to remove the bias delay use, but lead to an effective tool.
AI in dermatology is meant to create tools that are universally useful. To do this, developers need data featuring darker skin. Dermatology AI is trained using images to identify skin conditions like melanoma. If most images are lighter skin tones, as in some studies, the AI becomes biased. “In the application of AI to skin diagnosis, if a program is familiar with seeing melanoma on Caucasian skin, it may struggle considerably to identify the same on POC.” This issue is called “overfitting;” the AI struggles to recognize new images, including diverse skin tones. The more data AI uses with darker skin, the greater its application for dermatology. Much the same as practitioners, AI needs more information to improve. The data may also be too generalized.
The Fitzpatrick Skin Tone scale, or FST, is a potential limiting factor. Most studies use it to grade skin tone of patients. However, it was designed to analyze various skin tones’ responses to UV radiation. It was never intended for use in image-based AI. Between types IV and VI, the FST lumps together multiple shades of skin tone. This causes AI to struggle with tones in between the shades shown, most often Alaskan Indigenous or Pacific Islander groups.
Resources such as the Inclusive Skin Color Project and the community-driven Brown Skin Matters are compiling data from all across the Internet, and accepting new contributions regularly. If you want to do your part to add to their data, visit Brown Skin Matters and share any images you can. Thanks to these databases, a new scale has been created. The recently developed Monk Skin Tone Scale (MST) creates a much more inclusive picture. By representing a wider range of natural skin tone, the MST creates a better framework for AI learning and practitioners. Next comes a framework for ethical development, like TechQuity.
[caption: the Monk Skin Tone scale (top) shows multiple shades of skin tone previously excluded in the FST]
TechQuity is a recently coined term describing equal tech representation. According to the Journal of Healthcare for the Poor and Underserved: “…an anti-racism and pro-equity approach to the use of technology, or TechQuity, must:
- Address structural racism and discrimination to achieve a diverse workforce…
- Collect and track data that is representative of the concerns and needs of populations that face health inequities.
- Deploy data-driven and technology strategies to hold health institutions accountable for achieving equity (Rhee, Kyu, et al)…
”This may sound like a long road to health equity, but in today’s world, changes may evolve faster than ever. Collaboration is key. It’s time for the system to welcome any help it can get from their BIPOC patients; potentially even you.
How Technology is Transforming the Field
As it stands, AI diagnostics and tele-dermatology are supplementary. Their drawbacks prevent them from replacing doctors—for now. Using AI to help at specific steps, patients see great results. Users mostly prefer using a simple app to supplement their health care. Tele-dermatology and AI can ease health costs and improve accessibility. And, while image-based AI may not be ready yet, it’s progressing. Healthy.io is one AI tool that already helps. The app assesses liver disease risk using a doctor-provided kit. Much more convenient than a test at the GP! With data from projects like Brown Skin Matters, AI might make dermatology just as simple.
The Next Steps to Equal Healthcare
Though minority groups have been underrepresented, positive change is beginning. The gap in BIPOC skin care may soon be closed. For BIPOC patients, these tools are more than convenient—they’re a lifeline to early diagnosis and treatment. Mainstream healthcare embracing these solutions makes dermatology more available. As data improves, so will these tools. Currently, only 3% of dermatologists in the US are Black, and less than 5% are Hispanic or Latino. New initiatives are working to improve BIPOC representation, such as American Academy of Dermatology’s Pathways program. Practitioners’ training is also improving in key areas, improving BIPOC treatment. Simultaneously, AI is becoming a useful resource for all populations.
To ensure these technologies support equitable healthcare, developing AIs need extensive, inclusive data. Clinicians should be made aware of this problem and how to contribute to its solution. The Inclusive Skin Color Project is revolutionizing how dermatology addresses BIPOC skin conditions. Resources like this empower both AI development and clinician training. Think of it as the voice of the underserved, cutting through all the noise.
Every contribution counts. The future of equal healthcare may not be far off if professionals and patients work together. Early results show promising improvements. For example, training dermatology residents can work with over double efficiency through teledermatology (Gonzalez et. al). They gain more experience and help more people before they start their own clinics. So, even if you don’t have any skin condition, spread the word. Together, your community can help professionals create tools that work for everyone.
Sources
- Khatun, Nazma, et al. “Technology Innovation to Reduce Health Inequality in Skin Diagnosis and to Improve Patient Outcomes for People of Color: A Thematic Literature Review and Future Research Agenda.” Frontiers in Artificial Intelligence, U.S. National Library of Medicine, 13 June 2024, pmc.ncbi.nlm.nih.gov/articles/PMC11209749/.
- “Dermatology Guidelines Lack BIPOC-Specific Considerations.” Dermatology Advisor, Dermatology Advisor, 17 May 2024, www.dermatologyadvisor.com/news/derm-guidelines-lack-bipoc-specific-considerations/.
- Yousuf, Yusef, and Jaime C Yu. “Improving Representation of Skin of Color in a Medical School Preclerkship Dermatology Curriculum.” Medical Science Educator, U.S. National Library of Medicine, 30 Nov. 2021, pmc.ncbi.nlm.nih.gov/articles/PMC8631695/.
- Giansanti, Daniele. “Advancing Dermatological Care: A Comprehensive Narrative Review of Tele-Dermatology and mHealth for Bridging Gaps and Expanding Opportunities beyond the COVID-19 Pandemic.” Healthcare (Basel, Switzerland), U.S. National Library of Medicine, 1 July 2023, pmc.ncbi.nlm.nih.gov/articles/PMC10340283/.
- Sondermann W., Von Kalle C., Utikal J.S., Schadendorf D., Esser S., Durani B., Durani H., Jansen M., Brinker T.J. Externe wissenschaftliche Evaluation der ersten Teledermatologie-App ohne direkten Patientenkontakt in Deutschland (“Online Hautarzt–AppDoc”) [External scientific evaluation of the first teledermatology app without direct patient contact in Germany (Online Dermatologist-AppDoc)] Hautarzt. 2020;71:887–897. doi: 10.1007/s00105-020-04660-w.
- Rizvi S.M.H., Schopf T., Sangha A., Ulvin K., Gjersvik P. Teledermatology in Norway using a mobile phone app. PLoS ONE. 2020;15:e0232131. doi: 10.1371/journal.pone.0232131.
- “UCSF Guides: Anti-Racism Resources at UCSF Library: Inclusive Skin Color Project.” Inclusive Skin Color Project – Anti-Racism Resources at UCSF Library – UCSF Guides at University of California, San Francisco, guides.ucsf.edu/c.php?g=1081119&p=9159811. Accessed 3 Dec. 2024.
- Skin and Digital–the 2024 Narrative, www.mcpdigitalhealth.org/article/S2949-7612(24)00043-9/fulltext. Accessed 3 Dec. 2024.
- Maddi Hebebrand, MC. “Exploring the Perception of Digital Health Tools in Dermatology.” Dermatology Times, Dermatology Times, 6 Sept. 2024, www.dermatologytimes.com/view/exploring-the-perception-of-digital-health-tools-in-dermatology.
- Brinker T. J., Hekler A., Enk A. H., Klode J., Hauschild A., Berking C., et al. (2019). A convolutional neural network trained with Dermoscopic images performed on par with 145 dermatologists in a clinical melanoma image classification task. Eur. J. Cancer 111, 148–154. doi: 10.1016/j.ejca.2019.02.005
- Aggarwal P. (2019). Data augmentation in dermatology image recognition using machine learning. Skin Res. Technol. 25, 815–820. doi: 10.1111/srt.12726, PMID
- Obermeyer Z., Powers B., Vogeli C., Mullainthan S. (2019). Dissecting racial Bias in an algorithm used to manage the health of populations. Science 366, 447–453. doi: 10.1126/science.aax2342
- Rhee, Kyu, et al. “What Is TechQuity?” Journal of Health Care for the Poor and Underserved, Johns Hopkins University Press, 1 May 2021, muse.jhu.edu/article/789652.
- Harvard Gazette. “Teaching Algorithms about Skin Tones.” Harvard Gazette, 20 Dec. 2022, test.news.harvard.edu/gazette/story/2022/07/teaching-algorithms-about-colors-of-people/.
- Eedy D. (2015). Dermatology: a specialty in crisis. Clin. Med. 15, 509–510. doi: 10.7861/clinmedicine.15-6-509
- Shore J., Green M., Hardy A., Livesey D. (2019). The compliance and cost-effectiveness of smartphone urinalysis albumin screening for people with diabetes in England. Expert Rev. Pharmacoecon. Outcomes Res. 20, 387–395. doi: 10.1080/14737167.2019.1650024
- Gonzalez S;Syder N;McKenzie SA;Price KN;Ngo B;Mora Hurtado AC;Rodman J;Elbuluk N; “Racial Diversity in Academic Dermatology: A Cross-Sectional Analysis of Black Academic Dermatology Faculty in the United States.” Journal of the American Academy of Dermatology, U.S. National Library of Medicine, pubmed.ncbi.nlm.nih.gov/37734667/. Accessed 31 Dec. 2024.
- Boyers, Lindsay N, et al. “Teledermatology as an Educational Tool for Teaching Dermatology to Residents and Medical Students.” Telemedicine Journal and E-Health : The Official Journal of the American Telemedicine Association, U.S. National Library of Medicine, Apr. 2015, pmc.ncbi.nlm.nih.gov/articles/PMC4378857/. [19]E;, Zakaria A;Maurer T;Amerson. “Impact of Teledermatology Program on Dermatology Resident Experience and Education.” Telemedicine Journal and E-Health : The Official Journal of the American Telemedicine Association, U.S. National Library of Medicine, pubmed.ncbi.nlm.nih.gov/33217240/. Accessed 31 Dec. 2024.