Reshad Aquib Irteza, Nino Valentina, Valero Maria
Department of Industrial and Systems Engineering, Kennesaw State University, Marietta, GA, USA.
Department of Information Technology, Kennesaw State University, Marietta, GA, USA.
Vasc Health Risk Manag. 2025 Aug 30;21:685-703. doi: 10.2147/VHRM.S508620. eCollection 2025.
Cardiac arrhythmias are a major health concern around the world, causing morbidity and mortality in a wide range of people. The timely and accurate diagnosis of arrhythmias is critical for optimal clinical management and intervention. Deep learning techniques have developed as powerful tools for detecting arrhythmias in recent years, taking advantage of advances in signal processing and machine learning. This review investigates the use of deep learning approaches to detect arrhythmias via electrocardiogram (ECG) readings. The study includes an in-depth evaluation of 30 papers retrieved from three distinct databases using a structured method. The result indicates that deep learning models can achieve high accuracy like 99.93% as well as high F1 scores such as 99.57%. Furthermore, the study examines current research trends, approaches, and developments in deep learning-based arrhythmia detection, including convolutional neural networks (CNNs), and hybrid architectures that includes RNN and CNN algorithms. Additionally, the paper investigates the strengths and limits of existing techniques, focusing on critical issues such as dataset heterogeneity, model interpretability, and real-time implementation. Future research and development directions in arrhythmia detection using deep learning are also mentioned. This study seeks to give significant insights for physicians, researchers, and policymakers involved in the development and implementation of sophisticated arrhythmia detection systems, with the ultimate goal of improving patient outcomes and cardiac healthcare.
心律失常是全球主要的健康问题，在广泛人群中导致发病和死亡。心律失常的及时准确诊断对于优化临床管理和干预至关重要。近年来，深度学习技术借助信号处理和机器学习的进展，已发展成为检测心律失常的强大工具。本综述研究了通过心电图（ECG）读数使用深度学习方法检测心律失常的情况。该研究使用结构化方法对从三个不同数据库检索到的30篇论文进行了深入评估。结果表明，深度学习模型可以实现99.93%这样的高精度以及99.57%这样的高F1分数。此外，该研究考察了基于深度学习的心律失常检测的当前研究趋势、方法和进展，包括卷积神经网络（CNN）以及包括RNN和CNN算法的混合架构。此外，本文研究了现有技术的优势和局限性，重点关注数据集异质性、模型可解释性和实时实施等关键问题。还提到了使用深度学习进行心律失常检测的未来研发方向。本研究旨在为参与复杂心律失常检测系统开发和实施的医生、研究人员和政策制定者提供重要见解，最终目标是改善患者预后和心脏保健。
