INDIVIDUALIZED FEDERATED LEARNING FOR MULTI-CENTER INTENSIVE CARE UNIT HOSPITAL READINESS

Abstract: The healthcare industry is increasingly leveraging artificial intelligence (AI) to enhance patient outcomes, particularly in Intensive Care Units (ICUs) where timely and accurate predictions can save lives.[1] Personalized Federated Learning (PFL) offers a novel approach to this challenge by enabling collaborative learning across multiple centers while preserving patient privacy.[2] This research focuses on developing a PFL framework tailored for multi-center ICU prediction. The framework aims to integrate diverse patient data from various hospitals to predict critical outcomes, such as patient deterioration and mortality, without sharing sensitive data.[3] By personalizing models to local data characteristics while benefiting from the collective knowledge of all participating centers, the PFL approach is expected to improve prediction accuracy and patient care in ICUs.[4][5]

Keywords: Personalized Federated Learning, In-Hospital Mortality Prediction, Multi-Center ICU, Machine Learning, Healthcare Analytics, Federated Learning, ICU Patient Data, Mortality Risk Assessment