Datasets:

harvardairobotics
/

FairFedMed

license: mit
task_categories:
  - image-classification
modality:
  - image
language:
  - en
tags:
  - medical
  - ophthalmology
  - radiology
  - fairness
  - federated-learning
  - fundus
  - glaucoma
  - chest-xray
  - OCT
pretty_name: FairFedMed
size_categories:
  - 10K<n<100K

Dataset Card: FairFedMed

Dataset Summary

FairFedMed is the first federated learning (FL) benchmark dataset for medical imaging with demographic annotations, designed to study group fairness across institutions in a federated setting. It comprises two subsets spanning ophthalmology and chest radiology, enabling research on fairness-aware federated learning under realistic cross-institutional data heterogeneity.

This dataset was introduced in the IEEE Transactions on Medical Imaging 2025 paper: FairFedMed: Benchmarking Group Fairness in Federated Medical Imaging with FairLoRA.

Dataset Details

Dataset Description

Curated by: Minghan Li, Congcong Wen, Yu Tian, Min Shi, Yan Luo, Hao Huang, Yi Fang, Mengyu Wang
Institution: Harvard Medical School / Harvard AI and Robotics Lab
License: See individual subset licenses (CheXpert and MIMIC-CXR have their own terms)
Repository: Harvard-AI-and-Robotics-Lab/FairFedMed
Paper: IEEE TMI 2025 / arXiv:2508.00873

Subsets

FairFedMed-Oph (Ophthalmology)

Field	Value
Task	Glaucoma detection (binary classification)
Modalities	2D SLO fundus images, 3D OCT B-scans
Scale	15,165 patients
Demographics	Age, gender, race, ethnicity, preferred language, marital status (6 attributes)
FL Setup	Multi-site federated (3 sites)

FairFedMed-Chest (Chest Radiology)

Field	Value
Task	Chest pathology classification
Sources	CheXpert + MIMIC-CXR
Demographics	Age, gender, race (3 attributes)
FL Setup	2 clients simulating cross-institutional FL

Uses

Direct Use

Research on group fairness in federated medical image classification, including studies of demographic disparity across institutions and evaluation of fairness-aware FL methods.

Out-of-Scope Use

Clinical diagnosis, commercial applications. Note that FairFedMed-Chest inherits the usage restrictions of CheXpert and MIMIC-CXR — consult those datasets' licenses before use.

Evaluation

Metric	Description
AUC	Area Under ROC Curve
ESAUC	Equalized Selection AUC
EOD	Equalized Odds Difference
SPD	Statistical Parity Difference
Group AUC	Per-demographic-group AUC

Associated Method: FairLoRA

The paper introduces FairLoRA, a fairness-aware FL framework using SVD-based low-rank adaptation. It customizes singular values per demographic group while sharing singular vectors across clients for communication efficiency.

Supported backbones: ViT-B/16, ResNet-50.

Citation

BibTeX:

@ARTICLE{11205878,
  author={Li, Minghan and Wen, Congcong and Tian, Yu and Shi, Min and Luo, Yan and Huang, Hao and Fang, Yi and Wang, Mengyu},
  journal={IEEE Transactions on Medical Imaging},
  title={FairFedMed: Benchmarking Group Fairness in Federated Medical Imaging with FairLoRA},
  year={2025},
  pages={1-1},
  doi={10.1109/TMI.2025.3622522}
}

APA:

Li, M., Wen, C., Tian, Y., Shi, M., Luo, Y., Huang, H., Fang, Y., & Wang, M. (2025). FairFedMed: Benchmarking Group Fairness in Federated Medical Imaging with FairLoRA. IEEE Transactions on Medical Imaging. https://doi.org/10.1109/TMI.2025.3622522