Tschandl P, et al. - Researchers assessed the accuracy of machine-learning algorithms in diagnosing all pigmented skin lesions vs human experts in this open, web-based, international, diagnostic study. They compared the diagnoses from human readers with those of 139 algorithms created by 77 machine-learning labs, who participated in the International Skin Imaging Collaboration 2018 challenge and received a training set of 10,015 images in advance. From 63 countries, 511 human readers comprising 283 (55.4%) board-certified dermatologists, 118 (23.1%) dermatology residents, and 83 (16.2%) general practitioners had a minimum of one attempt in the reader study. Compared to human experts, state-of-the-art machine-learning classifiers performed better in the diagnosis of pigmented skin lesions and should have a more important role in clinical practice. However, these algorithms had decreased performance for out-of-distribution images; this should be investigated in future work.