CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep\n Learning

Abstract

We develop an algorithm that can detect pneumonia from chest X-rays at a\nlevel exceeding practicing radiologists. Our algorithm, CheXNet, is a 121-layer\nconvolutional neural network trained on ChestX-ray14, currently the largest\npublicly available chest X-ray dataset, containing over 100,000 frontal-view\nX-ray images with 14 diseases. Four practicing academic radiologists annotate a\ntest set, on which we compare the performance of CheXNet to that of\nradiologists. We find that CheXNet exceeds average radiologist performance on\nthe F1 metric. We extend CheXNet to detect all 14 diseases in ChestX-ray14 and\nachieve state of the art results on all 14 diseases.\n