AI-Based Real-Time ECG Anomaly Detection Using Temporal Convolutional Networks: A Comprehensive Review

Authors:

^1*Aman shrivastava
Research Scholar, Department of Computer Science, Technocrats Institute of Technology and Science Bhopal, India
^2*Dr Mukesh Asati

Assistant Professor, Department of Computer Science, Technocrats Institute of Technology and Science, Bhopal

^3*Prof Rakesh Kumar Tiwari

Assistant Professor, Department of Computer Science, Technocrats Institute of Technology and Science, Bhopal

^4*Prof Ekta Bisen

Assistant Professor, Department of Computer Science, Technocrats Institute of Technology and Science, Bhopal

Abstract

Cardiovascular diseases remain a major cause of death in the world, and thus it is imperative to have correct and timely diagnoses of cardiac abnormalities. The direct interpretation of the electrocardiogram (ECG) signals is manual and is time consuming and prone to human error, especially when there is constant monitoring. The following paper is a review of a Temporal Convolutional Networks (TCN)-based real-time ECG anomaly detection framework using AI. The system reviewed has ECG signals obtained in the MIT-BIH Arrhythmia Database and applies noise reduction and heartbeat segmentation as preprocessing procedures to enhance the quality of the signal. It uses a powerful multi-feature extraction framework based on Fractional Discrete Cosine Transform (FrDCT), Radon Wavelet Transform (RWT) and Fractional Wavelet Transform (FrWT) to ensure that the ECG data is represented in a holistic manner. A Concrete Autoencoder (CAE) is used to reduce the dimensions by picking out the most significant features. TL-MobileNet-ED A new Transformer-based MobileNet Encoder-Decoder that incorporates Temporal Convolutional Networks (TL-MobileNet-ED) is proposed to perform efficient and accurate anomaly classification. Experimental findings indicate that the classification accuracy is 98.43 with a matching high precision, recall and F1-score as compared to current machine learning and deep learning benchmarks. The paper summarizes the methodology, presents the analysis of the experimental findings, and explains the general implications of the framework to real-time intelligent cardiac monitoring.

Keywords: Electrocardiogram (ECG), ECG Anomaly Detection, Artificial Intelligence (AI), Deep Learning, Temporal Convolutional Networks (TCN), MobileNet, Transformer Learning, Feature Extraction, Concrete Autoencoder (CAE), Real-Time Monitoring.