Deep Learning Approaches for Atrial Fibrillation Detection Using Single-Lead ECG Signal Analysis: A Review

Abstract

Atrial fibrillation (AF) is one of the most prevalent cardiac arrhythmias and a major risk factor for stroke and heart failure, necessitating accurate and early detection. Recent advancements in artificial intelligence, particularly deep learning, have significantly improved AF detection using single-lead electrocardiogram (ECG) signals. This paper presents a comprehensive review of deep learning and optimization approaches that enhance AF detection accuracy through hybrid signal processing techniques, including Fast Fourier Transform (FFT) and Continuous Wavelet Transform (CWT), combined with stochastic pooling neural network architectures. The integration of time-frequency representations enables efficient extraction of both spectral and temporal features from non-stationary ECG signals. Deep learning models such as convolutional neural networks (CNN), recurrent neural networks (RNN), and hybrid CNN-LSTM architectures demonstrate superior performance compared to traditional methods. Studies from 2020–2023 show that combining raw ECG signals with transformed features significantly improves classification accuracy and robustness. Optimization techniques further enhance convergence and generalization. However, challenges such as noise, data imbalance, and computational complexity persist. This review analyzes recent literature, compares methodologies, and identifies future research directions for developing efficient and scalable AF detection systems.