<p>Healthcare blockchains must support real-time clinical workloads, varied network nodes, and high privacy. The proposed architecture includes QoS-driven sharding, behavior-based governance, semantic transaction priority, adaptive role allocation, and verifiable federated learning. Multi-dimensional performance measures reduce shard formation, learning-based trust evaluation improves consensus integrity, and a transformer-based urgency classifier speeds up medical transactions. Dynamic role-management improves energy efficiency, and a privacy-preserving federated learning layer permits cross-institutional model development without revealing local data samples. In a heterogeneous 60-node environment with clinical datasets, latency, shard balancing, trust accuracy, semantic classification, and learning performance improves for different scenarios. The results demonstrate that a blockchain ecosystem with contextual information, adaptive governance, and encrypted collaborative analytics can support large-scale healthcare operations. Security, efficiency, and clinical awareness are embedded into decentralized healthcare infrastructures in process. This integrated architecture reduces transaction latency by 12.8%, load balancing by 17.4%, and malicious nodes by 21.5% while ensuring 100% data privacy. Improvements make the proposed system a high-impact, scalable solution for intelligent and secure healthcare blockchain systems.</p>