Agentic AI in Healthcare: A Comprehensive Survey of Foundations, Taxonomy, and Applications

Le résumé fourni par la source

Agentic AI marks a paradigm shift in healthcare, moving beyond predictive models toward autonomous, goaldirected systems capable of perceiving, reasoning, acting, and adapting in dynamic clinical contexts. Earlier generations of AI in healthcare focused on narrow tasks such as disease classification, image analysis, and structured data prediction. While impactful, these approaches were constrained by limited adaptability, lack of memory, and inability to coordinate across complex workflows. In contrast, agentic AI integrates multimodal data from electronic health records, imaging, wearables, and patientreported outcomes, applies contextual reasoning, leverages tool use and APIs, and incorporates memory and feedback loops to support longitudinal, personalized care. This transition offers significant opportunities. Agentic systems can streamline clinical workflows, augment decision support, automate documentation, manage resources, and engage patients through interactive and adaptive interfaces. Multi-agent architectures further enable distributed collaboration, where specialized agents coordinate across domains such as radiology, oncology, and emergency care, resembling real-world clinical teams. By shifting from static predictions to continuous sense-think-act cycles, agentic AI has the potential to deliver more responsive, personalized, and proactive healthcare. However, increased autonomy also raises critical challenges. Issues of safety, transparency, bias, accountability, and interoperability remain barriers to clinical integration. Without robust evaluation, governance frameworks, and human oversight, these systems risk over-reliance, propagation of errors, and ethical concerns around patient trust and privacy. Addressing these challenges requires rigorous validation, domain adaptation, and mechanisms for safe collaboration between humans and AI agents. The goal of this paper is to provide a comprehensive survey of agentic AI in healthcare, covering conceptual foundations, taxonomies, enabling technologies, architectures, and applications across clinical, operational, and research domains. We also highlight opportunities, limitations, and open research directions to guide the responsible development of safe, secure, and scalable agentic systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.