CheXpert

CheXpert is a large public dataset for chest radiograph interpretation, consisting of 224,316 chest radiographs of 65,240 patients collected from Stanford Hospital.

Key Features

• Scale: Over 200,000 images, making it one of the largest publicly available collections of annotated X-rays.
• Automated Labeling: Labels were automatically extracted from radiology reports using a natural language processing labeler.
• 14 Observations: The dataset is labeled for 14 common chest radiographic observations (e.g., Pneumonia, Cardiomegaly, Pleural Effusion).
• Uncertainty Labels: Unique among datasets, it explicitly captures uncertainty in radiology reports, allowing models to learn from ambiguity.

Getting Started

For guidance on how to build a deep learning classifier using the CheXpert dataset and PyTorch, see Cookbook 19: Training Chest X-ray Classifiers in our Getting Started guide.