AI-Based Federated Learning for Rare Hematological Diseases

Authors: Asti G, D'Amico S, Merleau NSC, et al.

Published in: Blood (2024)

Citations: 9

Abstract:
Rare hematological diseases suffer from small patient cohorts at individual centers, limiting statistical power for AI model training. This paper presents a federated learning framework that allows multiple hospitals to collaboratively train predictive models without sharing raw patient data. Applied to precision medicine tasks in rare blood cancers, the platform achieves performance comparable to centralized training while preserving patient privacy, offering a practical path to AI-assisted diagnosis in data-scarce medical domains.