Current AI programs do worse at identifying skin lesions in people of color, research shows

Artificial intelligence (AI) programs have recently acquired widespread popularity in dermatology for assessing, diagnosing, and treating skin conditions. As reported in the International Journal of Dermatology, investigators recently analyzed all published studies from the last 10 years to evaluate current AI programs in use for dermatologic purposes, uncovering significant shortcomings when applied to skin of color (SOC).

The researchers identified various challenges when this technology is applied to SOC, mainly stemming from the underrepresentation of SOC in datasets and issues with image quality and standardization. Results indicate that current AI programs inevitably do worse at identifying skin lesions in SOC. Also, only 30% of the programs identified in this analysis had data specifically in SOC.

The idea for this research project stemmed from lead author Rebecca Fliorent's establishment of South Jersey Skin Talk, an organization committed to educating others about dermatologic conditions, with an emphasis on those affecting people of color. Inspired by the profound absence of images portraying skin of color, Rebecca investigated the role of AI in dermatology, particularly its implications for addressing the needs of diverse skin types." 

Thu Minh Truong, PharmD, RPh, corresponding author, medical student at Rutgers New Jersey Medical School