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

CheXNet is an algorithm developed to detect pneumonia from chest X-rays at a level exceeding practicing radiologists. It is a 121-layer convolutional neural network trained on ChestX-ray14, at the time the largest publicly available chest X-ray dataset, containing over 100,000 frontal-view X-ray images labeled with 14 diseases. To evaluate performance against clinical experts, four practicing academic radiologists annotated a test set, on which the authors directly compared CheXNet to the radiologists.

The study finds that CheXNet exceeds average radiologist performance on the F1 metric for pneumonia detection. The authors further extend CheXNet to detect all 14 diseases present in ChestX-ray14 and achieve state-of-the-art results on all 14 disease categories. This demonstration that a deep network could match or surpass radiologists on a well-defined chest X-ray task became an influential example of deep learning applied to medical imaging.