The integration of Artificial Intelligence (AI) with Electronic Health Records (EHRs) provides immense potential in improving diagnostic, predictive analytics, and patient management in modern healthcare. However, centralizing healthcare data raises privacy concerns, trust issues, and regulatory challenges. This paper introduces a hybrid model combining Federated Learning (FL) and Blockchain technology to offer a decentralized, privacy-preserving approach for AI model training across multiple healthcare providers. The proposed framework uses smart contracts for adaptively dynamic patient consent and deploys adaptive optimization techniques, including asynchronous updates and gradient sparsification, while maintaining tamper-proof auditability via Blockchain. We evaluate our system in a multi-node cloud setting (Amazon Web Services Elastic Compute Cloud [AWS EC2]) using the Medical Information Mart for Intensive Care III (MIMIC-III) dataset and show that our model behaves similarly to centralized learning, having increased privacy and lowered communication overhead. The hybrid framework ensures responsible use of healthcare data following established Health Insurance Portability and Accountability Act (HIPAA) and General Data Protection Regulation (GDPR) practices, bridging the gap between technological advancement and healthcare. The architecture lays the foundation for scalable, secure, and collaborative digital health ecosystems, promoting patient trust and regulatory compliance.