204P Comparative analysis of a multiclass convolutional neural network and 96 dermatologists in skin lesion diagnosis: Findings from an international web-based study

Abstract

Artificial intelligence (AI) tools have demonstrated the ability to enhance diagnostic accuracy in skin cancer screenings. While most systems provide binary “benign/malignant” classifications, multiclass models may provide greater clinical utility as they allow a managerial triage of patients. Considering the shortage of board-certified dermatologists, this is of particular relevance for non-specialist health care providers performing skin cancer screening. However, comparisons between multiclass convolutional neural networks (CNNs) and dermatologist performance remain scarce.

Ähnliche Arbeiten