Explainable Federated Learning for Privacy-Preserving Diabetes Prediction
Combining privacy and explainability to enable secure, trustworthy AI in healthcare.
Overview
This research introduces a federated learning framework for diabetes prediction, allowing multiple institutions to collaborate without sharing sensitive patient data.
?The Challenge
Healthcare data is highly sensitive, making centralised AI training difficult. At the same time, many models lack transparency, reducing trust among clinicians and organisations.
✓Our Solution
We developed a federated learning system with integrated explainability, enabling distributed model training while preserving data privacy and providing interpretable predictions.
Key Features
Results & Impact
- 1Published in ACM ICCES 2026
- 2Maintained strong predictive performance across distributed data
- 3Enhanced trust through explainability
- 4Supports privacy-first healthcare AI adoption
Technical Approach
Federated learning architecture with local model updates aggregated centrally. Explainability applied using feature attribution methods across distributed models.
Future Directions
Expansion to multi-hospital deployments and integration with real-world healthcare systems.
Technologies Used
Interested in similar work?
We're always open to discussing new partnerships and research collaborations.
Get in touch