The integration of Artificial Intelligence (AI) and the Internet of Things (IoT) presents a transformative opportunity for the continuous monitoring and early detection of medical conditions. In this paper, we investigate the application of AI algorithms for cardiovascular disease detection through the statistical analysis of IoT data streams. Utilizing connected wearable devices for healthcare analytics introduces severe challenges, including erratic data quality, dynamic shifts in physiological baseline metrics, and stringent data privacy requirements. To address these multifaceted issues, we propose an adaptive, ensemble-based machine learning framework that continuously adjusts to physiological concept drift while operating within a secure, hardware-level trusted execution environment. By combining data trust evaluation methods with encrypted cloud processing, the proposed system ensures both the diagnostic integrity and the confidentiality of patient cardiovascular data. A hypothetical evaluation plan is outlined to demonstrate how the framework can maintain high accuracy and low latency when subjected to simulated, large-scale medical data streams.