Disparities in Dermatology AI: Assessments Using Diverse Clinical Images
This addresses biases in dermatology AI that could affect billions of people lacking access to care, but it is incremental as it identifies weaknesses without proposing a new solution.
The study tackled the problem of AI diagnostic tools for skin diseases performing poorly on diverse skin tones and uncommon diseases by curating the Diverse Dermatology Images (DDI) dataset, showing that state-of-the-art models' ROC-AUC dropped 29-40% on this dataset.
More than 3 billion people lack access to care for skin disease. AI diagnostic tools may aid in early skin cancer detection; however most models have not been assessed on images of diverse skin tones or uncommon diseases. To address this, we curated the Diverse Dermatology Images (DDI) dataset - the first publicly available, pathologically confirmed images featuring diverse skin tones. We show that state-of-the-art dermatology AI models perform substantially worse on DDI, with ROC-AUC dropping 29-40 percent compared to the models' original results. We find that dark skin tones and uncommon diseases, which are well represented in the DDI dataset, lead to performance drop-offs. Additionally, we show that state-of-the-art robust training methods cannot correct for these biases without diverse training data. Our findings identify important weaknesses and biases in dermatology AI that need to be addressed to ensure reliable application to diverse patients and across all disease.