Transforming Smart Healthcare Systems with AI-Driven Edge Computing for Distributed IoMT Networks

TL;DR

This paper proposes an AI-driven edge computing model to improve smart healthcare systems by efficiently managing distributed IoMT networks and reducing system overhead.

Contribution

The novel contribution is an AI-driven edge computing model that dynamically detects network anomalies in IoMT with minimal overhead.

Findings

The model reduces energy consumption by 53% and latency by 46% compared to existing solutions.

It also lowers packet loss rate by 52%, increases network throughput by 56%, and reduces overhead by 48%.

The model was verified through simulations using synthetic data.

Abstract

The Internet of Medical Things (IoMT) with edge computing provides opportunities for the rapid growth and development of a smart healthcare system (SHM). It consists of wearable sensors, physical objects, and electronic devices that collect health data, perform local processing, and later forward it to a cloud platform for further analysis. Most existing approaches focus on diagnosing health conditions and reporting them to medical experts for personalized treatment. However, they overlook the need to provide dynamic approaches to address the unpredictable nature of the healthcare system, which relies on public infrastructure that all connected devices can access. Furthermore, the rapid processing of health data on constrained devices often leads to uneven load distribution and affects the system’s responsiveness in critical circumstances. Our research study proposes a model based on AI-driven and edge computing technologies to provide a lightweight and innovative healthcare system. It enhances the learning capabilities of the system and efficiently detects network anomalies in a distributed IoMT network, without incurring additional overhead on a bounded system. The proposed model is verified and tested through simulations using synthetic data, and the obtained results prove its efficacy in terms of energy consumption by 53%, latency by 46%, packet loss rate by 52%, network throughput by 56%, and overhead by 48% than related solutions.

Figures