healthcaredatabase.org
Directory / Open datasets / HDB-0050

CheXpert

Stanford ML Group
HDB-0050

CheXpert is a large dataset comprising 224,316 chest radiographs from 65,240 patients collected from Stanford Hospital, with labels for 14 clinical observations extracted from radiology reports. The dataset is designed for developing and benchmarking algorithms for automated chest radiograph interpretation.

Category
Open dataset
Access
Free registration
Region
US
Size
224,316 images
License / terms
Stanford University Dataset Research Use Agreement
Data types
imaging, radiology, text-reports
chest-xrayradiologyStanfordmachine-learning
Visit official site ↗

Record HDB-0050 · last verified 2026-08-05 · listed in Open datasets

Related records

HDB-0001

MIMIC-IV

MIT Laboratory for Computational Physiology De-identified electronic health records from patients admitted to Beth Israel Deaconess Medical Center, covering ICU and emergency department stays from 2008 onward. The de facto standard dataset for clinical machine learning research, with vitals, labs, medications, procedures and linked clinical notes modules. Open dataset Credentialed HDB-0002

eICU Collaborative Research Database

MIT LCP / Philips Healthcare Multi-center ICU database from the Philips eICU telehealth program, covering 200,000+ admissions across hospitals in the United States. Complements MIMIC with multi-center generalizability for critical care research. Open dataset Credentialed HDB-0003

NHANES

CDC / National Center for Health Statistics The National Health and Nutrition Examination Survey combines interviews, physical examinations and laboratory tests on a representative sample of the US population, released in continuous two-year cycles since 1999. A workhorse dataset for epidemiology and public health research. Open dataset Open access HDB-0004

SEER Cancer Registry

National Cancer Institute The Surveillance, Epidemiology, and End Results program collects cancer incidence, treatment and survival data from population-based registries covering roughly half of the US population. The reference source for US cancer statistics. Open dataset By application