> ML_DATASET // CHEXPERT-CHEST-RADIOGRAPHS_v1.0
CheXpert: A Large Chest Radiograph Dataset (Irvin et al. 2019)
Stanford ML Group (Stanford Hospital) · Medical Imaging & Radiology · 224,316 chest radiographs of 65,240 patients with radiologist report extractions
Medical Imaging & RadiologyStanford University CheXpert Non-Commercial Research License224,316 chest radiographs of 65,240 patients with radiologist report extractionsregistration-required
Dataset Profile & Characteristics
Label Type:14 thoracic pathology observation labels (positive, negative, uncertain)
Languages:en
License Tier:non-commercial
Modalities:image, text
Intended Use
- Chest X-ray thoracic disease classification benchmarking
- Uncertainty-aware label modeling in medical imaging
Prohibited / Discouraged Use
- Autonomous clinical diagnostic reading without licensed radiologist review
Bias, Leakage & Privacy Risk Analysis
Privacy / Sensitive Data Risks:
De-identified radiographic images; radiologist reports cleaned of patient identifiers.
Known Bias:
Single healthcare system patient population; frontal vs lateral projection variance.
Known Benchmark Leakage:
Validation set evaluated against consensus of 3 board-certified radiologists.