The integration of Artificial Intelligence (AI) with the Internet of Medical Things (IoMT) has birthed a new generation of Clinical Decision Support Systems (CDSS) capable of real-time physiological monitoring. This review article examines the architectural and methodological shift from rule-based alerts to predictive AI engines that process high-frequency data from connected medical devices. We investigate the core pipeline of these systems—from signal denoising at the Edge to deep learning-based feature extraction in the Cloud—and evaluate how these technologies address the "data deluge" currently overwhelming clinical staff. The article provides a detailed taxonomy of AI methodologies, including Supervised Learning for diagnosis, Reinforcement Learning for treatment optimization, and the rising role of Explainable AI (XAI) in fostering clinician trust. Key clinical use cases are explored, ranging from early sepsis detection in the ICU to the management of chronic conditions like diabetes through closed-loop artificial pancreas systems. Furthermore, we address the critical barriers to adoption, specifically focusing on data quality, clinical alarm fatigue, and the "interoperability gap" between siloed medical systems. Finally, the review analyzes the 2025 regulatory landscape, including the impact of the EU AI Act and the FDA's evolving SaMD guidelines. We conclude that while AI-powered CDSS offers unprecedented potential for proactive care, its success depends on maintaining a "Human-in-the-Loop" approach, ensuring that AI augments rather than replaces clinical expertise.