Fig. 1: The deep-learning system.
From: Assessment of a deep-learning system for fracture detection in musculoskeletal radiographs
a The deep-learning system used an ensemble of 10 convolutional neural networks. To produce a prediction, radiographs are processed by each network in the ensemble, averaged, and then post-processed to generate an overall fracture determination and bounding boxes. b Example outputs for each of the 16 anatomical regions supported by the deep-learning system.
