Skip to main content

AI to reduce high skin cancer deaths in people of colour

POC920
Image: Shutterstock

People of colour (POC) are more likely to die from melanoma — a form of skin cancer —compared to white people, due to delayed diagnosis. In a new study, researchers at Brunel University London have explored how artificial intelligence (AI) can be used to improve skin cancer diagnosis in POC and increase survival rates.

As September signals the end of summer and the start of a new season, people are encouraged to do a post-summer skin check to observe any changes to existing moles or new spots after being in the sun. Skin cancer can affect us all, but POC are at a greater disadvantage in melanoma mortality rates due to late diagnosis or incorrect treatment.

“Although socioeconomic status and limited access to healthcare services can lead to health inequalities — unjust and avoidable differences in people's health across the population and between specific population groups — evidence demonstrates a notable disadvantage for POC in various aspects of healthcare, and misdiagnosis and ill-prepared practitioners can also be part of the problem,” explained Dr Federico Colecchia, a technology expert from Brunel University London, who is part of the research team.

Miss Nazma Khatun, a doctoral researcher from Brunel who works on the study, explained how the lack of medical insight and knowledge on skin lesion diagnosis in POC is a persistent issue. “Inadequate resources detailing symptoms on brown and black skin pose a challenge to accurately describe how symptoms appear on diverse skin tones, which slows down the creation of inclusive educational resources for clinical staff and trainees,” she said. “These inequalities can result in a poorer quality of life for POC, in comparison to white people, and a disproportionate under-representation of POC in medical advancements.”

AI is a fast-growing technology that enables machines to simulate human intelligence, and AI-powered tools have been created to increase the availability of expert skin information. Existing applications assess uploaded pictures of skin lesions to produce a diagnosis, and the Brunel researchers have explored how AI could be used to address health inequalities, benefiting both healthcare clinicians and POC.

“Having access to AI-powered digital technology remotely or in community healthcare settings can significantly reduce health inequalities by removing the challenges of costs, geographical location and transportation,” explained Miss Khatun.

Although AI has demonstrated promising outcomes in providing accurate diagnoses of skin conditions and addressing health inequalities, the researchers have highlighted key areas for development.

“An AI skin recognition tool may assist in diagnosing skin concerns at home or in a clinical setting by collecting relevant images and descriptions, but before AI can achieve a reliable diagnosis, several barriers must be overcome,” explained Dr Gabriella Spinelli, an expert in healthcare technology from Brunel, who supervises Miss Khatun and works on the study.

“These include the lack of POC in clinical research, the low volume of dermatology images from POC and even the classification of skin types, which is ambiguous at best,” she added.

“The capability of AI to match dermatologists' performance in terms of correct classification of skin lesions is achievable, in principle, but the fundamental issue of a lack of balanced datasets representative of all skin types is a gap that needs addressing for both traditional and AI-assisted diagnostic pathways,” explained Dr Colecchia.

“The scarcity of diverse images can have a significant impact on the reliability and generalisability of AI models if they are predominantly trained on lighter skin tones, and additional research and measures are essential to address variations in skin tone.”

Dr Spinelli reinforced that researchers and designers must guarantee the inclusion of a diverse and representative group of participants in studies and AI developments to address the unintentional biases within healthcare.

“Diversity in image datasets is crucial to prevent biases, and we must reinforce the need for more care to be placed in ensuring POC are being cared for at the same pace and level as white people,” she said. The research team is now completing an inclusive data collection using participants with different skin types.

Technology innovation to reduce health inequality in skin diagnosis and to improve patient outcomes for people of colour: A thematic literature review and future research agenda’, by Nazma Khatun, Gabriella Spinelli and Federico Colecchia, is published in Frontiers of Artificial Intelligence. 

Reported by:

Nadine Palmer, Media Relations
+44 (0)1895 267090
nadine.palmer@brunel.ac.uk