The rapid growth of the Internet of Medical Things (IoMT) and Artificial Intelligence (AI) in healthcare has created a need for secure and privacy-preserving collaborative learning systems. Federated Learning (FL) allows multiple healthcare institutions to train machine learning models without sharing raw data, but it is vulnerable to poisoning attacks where malicious participants send harmful model updates. To address this issue, this paper proposes a blockchain-based Federated Learning framework integrated with Secure Multi-Party Computation (SMPC) for secure model verification and aggregation. In the proposed system, local model updates are verified through an encrypted process to detect and remove poisoned updates before aggregation. The verified updates are then securely aggregated and stored using blockchain technology, ensuring decentralization, transparency, and tamper-proof records. Experimental results on medical datasets show that the proposed approach effectively identifies malicious updates, preserves data privacy, and maintains high model accuracy. Compared to traditional methods, the system provides better security and reliable collaborative learning with minimal impact on performance. This framework offers a robust and trustworthy solution for secure healthcare data analysis and can also be extended to other privacy-sensitive domains