Edge Artificial Intelligence for Real-Time Healthcare Monitoring: A Layered Framework for Low Latency, Privacy Preserving Patient Care

Abstract

The growing burden of chronic disease and the structural limitations of centralized cloud computing have intensified interest in edge based artificial intelligence as a mechanism for delivering continuous, low latency patient monitoring outside conventional clinical settings. This paper develops and critically examines a layered Edge AI framework that distributes inference, communication, and privacy preservation across sensing devices, local gateways, and coordinating cloud infrastructure rather than relying on a single centralized processing node. Drawing on a systematic synthesis of peer reviewed literature spanning edge computing, TinyML, federated learning, and Internet of Medical Things research, the study identifies recurring architectural patterns and unresolved tensions among diagnostic accuracy, energy consumption, and regulatory compliance. A qualitative, theory building methodology is adopted, combining conceptual analysis with a comparative synthesis of empirically validated systems reported in the literature between 2016 and 2026. The findings indicate that hybrid architectures combining convolutional and recurrent neural components, quantized for microcontroller class hardware, achieve latency reductions exceeding fifty percent relative to cloud dependent baselines while incurring only modest accuracy trade offs. Federated learning combined with homomorphic encryption further reduces communication overhead and strengthens compliance with data protection regulation, albeit at a measurable computational cost. The paper contributes a consolidated conceptual model that integrates these strands into a six layer reference architecture and offers implications for clinical deployment, health system policy, and future research on explainability and trust in autonomous monitoring systems.