Federated Learning (FL) has emerged as a transformative paradigm in collaborative machine learning, enabling decentralized training across distributed devices while preserving data privacy. Despite its promising potential, FL encounters notable challenges, including non-IID (non-independent and identically distributed) data distributions, communication bottlenecks, and computational limitations. This work explores the integration of advanced machine learning techniques within the FL framework, specifically in the healthcare domain, focusing on analyzing electrocardiogram (ECG) data to investigate apnea sleep periods. The findings reveal that, with proper implementation, FL can replace centralized methods, particularly in scenarios where data privacy and security are paramount.