PathologyFL

PathologyFL is the custom federated learning framework for computational pathology in this repository.

Goals:

privacy-preserving multi-site training
pathology-specific federated workflows
secure aggregation and differential privacy support
robustness to client heterogeneity and site imbalance
support for FAIR-WEIGHTS-H institutional weighting

Coordinator components

Typical coordinator services include:

orchestrator
aggregator
client registry
privacy engine
monitoring system
byzantine detection

Client-side workflow

text

Local pathology data
    -> local training
    -> update serialization
    -> secure communication
    -> aggregation

Validation status

Core federated integration tests pass and smoke tests execute on PCam-derived pathology data.

This validates end-to-end execution but not yet real multi-center clinical deployment.

For deeper configuration examples, see the legacy FL integration guide.