Evaluation of Artificial Intelligence-assisted Diagnosis of Skin Neoplasms - a single-center, paralleled, unmasked, randomized controlled trial.

A randomized trial (KCT0005614; https://cris.nih.go.kr) was conducted in a tertiary care institute in South Korea, to validate whether artificial intelligence (AI) could augment the accuracy of non-expert physicians in the real-world settings which included diverse out-of-distribution conditions. Four non-dermatology trainees and four dermatology residents examined the randomly allocated patients with skin lesions suspicious of skin cancer with or without the real-time assistance of AI algorithm (https://b2020.modelderm.com#world; convolutional neural networks). Using 576 consecutive cases (Fitzpatrick skin phototypes III or IV) with suspicious lesions out of the initial 603 recruitments, the accuracy of the AI-assisted group (n=295, 53.9%) was significantly higher than those of the Unaided group (n=281, 43.8%; P=0.019). The augmentation was more significant from 54.7% (n=150) to 30.7% (n=138; P<0.0001) in the non-dermatology trainees who had the least experience in dermatology. The augmentation was not significant in the dermatology residents. The algorithm could help the trainees in the AI-assisted group include more differential diagnoses than the Unaided group (2.09 diagnoses versus 1.95; P=0.0005). In this single-center, unmasked, paralleled, randomized controlled trial, the multiclass AI algorithm augmented the diagnostic accuracy of non-expert physicians in dermatology.