Optimizing AI Models for Haematological Malignancies with Federated Learning

Authors: Carota L, Casadei F, Asti G, Merleau NSC, et al.

Published in: Physica Medica (2026)

Citations: New

Abstract:
Federated learning for haematological malignancy diagnosis presents unique challenges due to class imbalance and heterogeneous data distributions across sites. This paper systematically optimizes AI model architectures and aggregation strategies for federated settings, combining simulation studies with real-world experiments to identify configurations that maximize diagnostic accuracy while minimizing communication rounds.