Bias in, bias out: Underreporting and underrepresentation of diverse skin types in machine learning research for skin cancer detection—A scoping review
Lisa N. Guo, Michelle S. Lee, Bina Kassamali, Carol Mita, Vinod E. Nambudiri
Abstract
To the Editor: Artificial intelligence (AI) and machine learning (ML) have the potential to improve and expand access to skin cancer screening. However, some populations may not benefit if technologies are developed with insufficiently diverse datasets.1 US cohorts used to train deep learning algorithms are disproportionately derived from a few states,2 and skin of color and racial/ethnic diversity represented in dermatologic AI applications are not well-characterized. We conducted a scoping review to investigate skin types and populations included in dermatologic ML research.
Topics & Concepts
MedicineSkin cancerCancer detectionDermatologyMachine learningCancerMedical physicsArtificial intelligenceInternal medicineComputer scienceCutaneous Melanoma Detection and ManagementSkin Protection and AgingNonmelanoma Skin Cancer Studies