DESIGNING A FEDERATED LEARNING FRAMEWORK FOR PRIVACY-PRESERVING MEDICAL IMAGING DIAGNOSTICS ACROSS DECENTRALIZED HOSPITAL SYSTEMS

The digitization of medical imaging and the growing interest in artificial intelligence (AI)-driven diagnostics have led to substantial innovations in healthcare analytics. However, traditional centralized learning methods compromise patient privacy and regulatory compliance. This paper proposes a federated learning (FL) framework that enables collaborative model training across multiple decentralized hospital systems without exchanging sensitive data. The framework enhances diagnostic model performance while ensuring data confidentiality and adhering to regulations such as HIPAA and GDPR. This study evaluates the efficacy, communication costs, and robustness of the framework using simulated datasets across distributed nodes.