Survival Model Optimization via Federated Learning

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

Published in: IEEE Big Data (2024)

Citations: 1 citations

Abstract:
Survival analysis — predicting time-to-event outcomes such as patient relapse — poses specific challenges for federated learning due to censored data and site-specific event rates. This paper adapts Cox proportional hazards models for federated optimization, evaluating convergence and prediction accuracy through simulation and multi-hospital experiments, achieving concordance indices within 2% of centralized baselines.