An ECG signal processing and cardiac disease prediction approach for IoT-based health monitoring system using optimized epistemic neural network.

Abstract

The rising prevalence of cardiac diseases necessitates advanced IoT-driven health monitoring systems for early detection and diagnosis. This study presents an efficient ECG-based cardiac disease prediction framework leveraging a multi-phase approach to enhance computational efficiency and classification accuracy. The Convolutional Lightweight Deep Auto-encoder Wiener Filter (CLDAWF) is employed for signal preprocessing, while the Quantized Discrete Haar Wavelet Transform (QD-HWT) extracts critical cardiac features, including P-wave fluctuations, QRS complex, and T-wave intervals. These refined features are classified using an optimized Epistemic Neural Network (ENN), whose parameters are fine-tuned via the Boosted Sooty Tern Optimization algorithm, improving accuracy and reducing system loss. The proposed model achieves 99.65% accuracy, demonstrating its effectiveness in real-time cardiac disease monitoring and offering a scalable, high-performance solution for IoT-based healthcare systems.