CheXpert: Chest X-Ray Images with Uncertainty Labels

CheXpert is a large dataset of chest X-rays for automated interpretation, featuring uncertainty labels and radiologist-labeled reference standard evaluation sets. The dataset was created by the Stanford ML Group and uploaded to Hugging Face by 'danjacobellis'. The Hugging Face record was last updated on July 18, 2024.

Use Cases

• Training computer vision models for automated chest X-ray interpretation based on the large collection of images.
• Developing models that handle diagnostic uncertainty based on the dataset's uncertainty labels.
• Benchmarking model performance against radiologist-labeled reference standard evaluation sets.

Strengths

• Dataset is described as 'large', suggesting a substantial scale for model training.
• Includes radiologist-labeled reference standard evaluation sets for benchmarking.
• Features uncertainty labels, which can be useful for training models to handle ambiguous cases.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The uploader notes a potential mapping issue with AP/PA labels, advising users to verify.

Provenance

Source

Stanford ML Group (https://stanfordmlgroup.github.io/competitions/chexpert/)

Collection Method

Likely collected from hospital systems for a competition on automated chest X-ray interpretation.

Freshness

Last updated 2024-07-18 14:11:06; freshness should be verified.