Q-BodyFogNets: A Novel Energy-Aware Deep Learning Framework for Predicting Heart Diseases in Fog-BAN Environments

Abstract

The Internet of Things (IoT), Fog, and Edge Computing are key innovations that have transformed body area networks (BAN) and their communication techniques. This combination of IoT with wireless networks has led to significant advancements, enabling energy and latency-aware BANs, which play a crucial role in smart healthcare applications. Recently, machine learning and deep learning algorithms have been applied to IoT-enabled BANs to address these limitations. However, the complexity of these learning algorithms can impact the overall network performance, and achieving both low power consumption and high diagnosis accuracy remains underdeveloped. This paper presents a novel, intelligent, and adaptive Q-learning framework for Fog-based body area networks (QAL-Fog-BAN) that handles incoming data from various dynamic BAN nodes. It deploys less complex recurrent neural networks (LCRNN) to achieve solutions such as energy-efficient and QoS-aware paths for medical data transmission. The paper also introduces a novel dataset, built using the ifOGSIM API and Python libraries. This approach collects network-centric parameters across seven attributes and 15,687 data points, which are used for evaluation and analysis of the proposed method's effectiveness in Fog-BAN networks. Comprehensive testing is performed with the collected data samples, and several metrics—such as accuracy, precision, recall, specificity, and F1-score—are assessed and compared with other intelligent Fog-BAN networks, including LEAF-NETS, FUELNETS, and WORN-DEAR. Simulation results demonstrate that the proposed design performs better than existing architectures and plays a major role in achieving lower energy consumption in Fog-BAN networks.