An Integrated AI-IoT Architecture for Real-Time Healthcare Monitoring and Intelligent Decision Support

Abstract

The increasing demand for continuous, low-latency patient monitoring has exposed significant vulnerabilities in traditional cloud-centric healthcare architectures, where milliseconds can determine clinical outcomes. This paper introduces the Adaptive SDN-Assisted Edge-Cloud Healthcare Intelligence (ASECHI) framework, a latency-aware AI-IoT architecture that combines edge computing, Software Defined Networking (SDN) and an Attention-enhanced Long Short-Term Memory (A-LSTM) model, for robust, real-time physiological analytics. This framework dynamically balances computation between edge and cloud layers using a novel orchestration mechanism that considers network latency, congestion state, edge resource availability, and patient criticality. The A-LSTM model captures non-linear temporal dependencies in multivariate physiological time-series while an attention mechanism selectively amplifies diagnostically significant time steps. ASECHI is evaluated on the MIMIC-III clinical database and achieves 97.1% classification accuracy.