Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

MIDRC Chest Radiographs for COVID (Lin 2024 subset)

Link

https://pmc.ncbi.nlm.nih.gov/articles/PMC11449211

Indexing

Keywords: COVID-19, chest radiograph, fairness, bias, race, sex, age, supervised contrastive learning

Content: CH

RadLex: RID10345, RID34769, RID5647

SNOMED: 840539006

Organization(s)

Medical Imaging and Data Resource Center (MIDRC)

American College of Radiology (ACR)

Radiological Society of North America (RSNA)

American Association of Physicists in Medicine (AAPM)

University of Chicago (hosting site for MIDRC repository)

Funding

MIDRC funded by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) under contracts 75N92020C00008 and 75N92020C00021; jointly led by ACR, RSNA, and AAPM.

Ethical review

Retrospective study approved by IRBs at each clinical center and Weill Cornell Medicine; informed consent waived due to publicly available nature of the datasets used.

Comments

Subset of MIDRC focused on chest radiographs for COVID-19 diagnosis with available age, sex, and race information; counts reflect data collected as of April 20, 2023.

Date

Updated: 2023-04-20

References

[1] Lin M, Li T, Sun Z, Holste G, Ding Y, Wang F, Shih G, Peng Y. "Improving Fairness of Automated Chest Radiograph Diagnosis by Contrastive Learning". Radiology: Artificial Intelligence. 2024-08-21. doi:10.1148/ryai.230342. PMID: 39166973. PMCID: PMC11449211.

Dataset

Motivation

To evaluate and mitigate algorithmic bias across demographic subgroups in automated COVID-19 diagnosis from chest radiographs.

Sampling

Included only computed radiography and digital radiography studies with available age, sex, and race information in MIDRC.

Partitioning scheme

Random patient-level split: 20% held-out test; remaining used for training/validation.

Relationships between instances

Multiple radiographs per imaging study and multiple studies per patient are present.

External data

Study also referenced NIH ChestX-ray14 for additional experiments and used MIMIC-CXR as an external test set.

Sensitive data

Demographic attributes (age, sex, race) are included for subgroup analyses.