Chest Radiographs in Congestive Heart Failure: Visualizing Neural Network Learning

Generative Visual Rationales can identify imaging features learned by a model trained to predict congestive heart failure from chest radiographs, allowing radiologists to better identify faults and biases.