An automated ECG signal categorization framework leveraging Long Short-Term Memory (LSTM) neural networks achieved a classification accuracy of 95-98% across four distinct rhythm classes.
An LSTM neural network framework can autonomously classify ECG signals into normal, atrial fibrillation, other irregular patterns, and noise with 95-98% accuracy, potentially aiding in automated arrhythmia detection.
Abstract: Electrocardiogram (ECG) recordings serve as a cornerstone of cardiac health surveillance, enabling clinicians to identify a broad spectrum of cardiovascular disorders, most notably arrhythmias such as atrial fibrillation (AF). Nevertheless, manual interpretation of these recordings is inherently laborious, technically demanding, and vulnerable to diagnostic inconsistencies owing to the sheer volume and heterogeneity of the data involved. This investigation proposes an innovative framework for fully automated ECG signal categorization, leveraging Long Short-Term Memory (LSTM) neural networks — a specialized class of deep learning architecture well-suited to the analysis of temporally evolving sequential data such as physiological waveforms. The methodology draws upon publicly available, annotated ECG repositories wherein individual signal segments are pre-labeled into four distinct classes: normal cardiac rhythm, atrial fibrillation, other irregular patterns, and noise or signal corruption. Unlike conventional pipelines, the LSTM-based architecture autonomously discovers discriminative temporal features within the raw waveform, rendering manual feature engineering unnecessary. Experimental validation demonstrates that the proposed system achieves a classification accuracy in the range of 95–98%, attesting to its clinical utility. Prospective deployment scenarios encompass wearable biosignal platforms, telehealth assessment environments, and AI-augmented clinical decision-support tools. Keywords: ECG, LSTM, Arrhythmia, Atrial Fibrillation, EMG Sensor, Artificial Intelligence, Deep Learning. Title: ECG Signal Classification Using Long Short-Term Memory (LSTM) Neural Networks Author: Dr. Jaya Raut, Asmita P., Kalyani P., Yashasvi R., Isha B., Reva K. International Journal of Novel Research in Electrical and Mechanical Engineering ISSN 2394-9678 Vol. 13, Issue 1, September 2025 - August 2026 Page No: 16-20 Novelty Journals Website: www.noveltyjournals.com Published Date: 20-April-2026 DOI: https://doi.org/10.5281/zenodo.19663520 Paper Download Link (Source) https://www.noveltyjournals.com/upload/paper/ECG%20Signal%20Classification%20Using%20Long-20042026-8.pdf
Raut et al. (Mon,) conducted a other in Arrhythmia and Atrial Fibrillation. Long Short-Term Memory (LSTM) neural networks was evaluated on Classification accuracy. An automated ECG signal categorization framework leveraging Long Short-Term Memory (LSTM) neural networks achieved a classification accuracy of 95-98% across four distinct rhythm classes.