Description

This PhD project aims to develop an Adaptive Causal Federated Continual Learning framework for personalized homecare monitoring. The proposed approach will enable AI models to learn collaboratively from decentralized and privacy-sensitive patient data, adapt continuously to changes in health conditions, devices, and lifestyles, and preserve previously acquired knowledge. By integrating causal representation learning, the framework will distinguish medically meaningful relationships from spurious correlations and provide interpretable explanations for clinical decision-making. The resulting system will be evaluated on temporal healthcare datasets and validated with clinical experts to ensure personalization, robustness, privacy compliance, and clinical relevance.