Detecting Cardiac Arrhythmias through Electrocardiography: Current Advancement and Future Direction from the Standpoint of Deep Learning

Abstract

Cardiac arrhythmia remains a leading cause of mortality worldwide and is a significant risk factor for the development of various cardiovascular diseases. Electrocardiography (ECG) is a widely utilized diagnostic tool for the early detection of cardiac arrhythmias, and recent advancements in deep learning (DL) have demonstrated notable success in automating and enhancing this process. Despite the growing body of research, there remains a lack of a focused and comprehensive literature review dedicated specifically to the application of deep learning techniques in ECG-based arrhythmia detection. Addressing this gap, the present study systematically reviews recent contributions that apply deep learning algorithms to ECG data for the identification and classification of cardiac arrhythmias. The review categorizes relevant studies based on architectural approaches, datasets used, performance metrics, and clinical relevance. A novel taxonomy is proposed to classify the domains of deep learning applications in ECG, including supervised, unsupervised, and hybrid learning models, as well as real-time and offline diagnostic systems. The review also identifies current limitations in model generalizability, data quality, and interpretability. Based on these insights, future research directions are proposed to guide the development of more robust, transparent, and clinically applicable deep learning systems for cardiac arrhythmia detection. This review serves as a foundational reference for researchers and practitioners seeking to advance the intersection of artificial intelligence and cardiovascular diagnostics.