Quote of the day

A person who never made a mistake never tried anything new.

- Albert Einstein

Contact

Fairness-Aware Federated Domain Generalization for Multi-Center Medical Imaging in Low-Resource Hospitals

Open Medicine & Healthcare

Fairness-Aware Federated Domain Generalization for Multi-Center Medical Imaging in Low-Resource Hospitals

Fairness-Aware Federated Domain Generalization for Multi-Center Medical Imaging in Low-Resource Hospitals

Description

Investigate how a medical-imaging model trained collaboratively across unequally equipped hospitals can remain accurate and fair for under-resourced sites.

Details

Context and Problem Statement

Federated learning is widely proposed for medical-data collaboration, yet many evaluations assume relatively homogeneous centers. In reality, large university hospitals may dominate aggregation while regional hospitals with older equipment remain underserved.

Research Question

How can federated aggregation improve performance at the worst-performing site without materially degrading average performance under strong equipment and protocol heterogeneity?

Proposed Approach

Compare fairness-aware aggregation, local personalization through adapters, image harmonization, and domain-generalization techniques. Evaluate per-site performance, inter-site disparity, and subgroup fairness when demographic information is available.

Expected Contribution

An empirical characterization of the fairness-performance trade-off under realistic federated medical-imaging heterogeneity.

Expected Prototype

A federated multi-site platform with a dashboard showing performance by center and subgroup.

Datasets

CheXpert, MIMIC-CXR, and BraTS partitioned to reproduce realistic center heterogeneity.

Challenges

Limited access to real multi-center data, reconstruction attacks, regulatory compliance, and contested definitions of clinical fairness.

Research Question

Does fairness-aware federated aggregation improve performance at disadvantaged medical-imaging sites, and what cost does this impose on global predictive performance?

Innovation

The project treats inter-site fairness as an optimization objective under realistic equipment inequality rather than a post-hoc metric.

Expected Deliverable

A reproducible federated benchmark, a quantitative fairness-performance analysis, and a conference paper in medical AI or biomedical informatics.

Technologies

Federated Learning (FedAvg, FedProx, Fair Aggregation) Domain Generalization and Image Harmonization Deep Learning for Medical Imaging Privacy-Preserving AI Algorithmic Fairness Metrics

Required Skills

  • Deep Learning and Medical Imaging
  • Federated Learning and Distributed Systems
  • Statistics and Fairness Evaluation
  • Health-Data Protection Frameworks

Datasets

  • CheXpert
  • MIMIC-CXR (PhysioNet, controlled access)
  • BraTS (multi-center segmentation)

Morocco & Africa Relevance

Hospital systems in Morocco and Africa often show strong disparities between university and provincial facilities; fair collaborative learning is therefore essential.

References

  1. Achieving flexible fairness metrics in federated medical imaging
  2. Advancements and challenges of federated learning in medical imaging: a systematic literature review
  3. Ethics and governance of artificial intelligence for health: large multi-modal models (WHO guidance)

Project Information

Duration: 4 to 6 months
Academic Level: Engineering, Master's or PhD
Research Area: Medicine & Healthcare

Keywords

Federated Learning Medical Imaging Fairness Domain Generalization Privacy

Evaluation

Scientific novelty
Impact
Technical feasibility
Data availability
Prototype potential
Industrial relevance

Apply for this topic

To apply for this topic, please send your CV and portfolio to:

contact@yanecode.com
Back to Research