A Comprehensive Review of Enhancing Atrial Fibrillation Detection Accuracy Based on Single Lead ECG Signal Analysis using Hybrid Fast Fourier and Continuous Wavelet Transforms and Stochastic Pooling Layer Neural Networks

Abstract

Atrial fibrillation (AF) is one of the most prevalent cardiac arrhythmias, significantly increasing the risk of stroke, heart failure, and mortality. Early and accurate detection using electrocardiogram (ECG) signals is crucial for timely diagnosis and treatment. Recent advances in signal processing and deep learning have enabled automated AF detection systems, particularly using single-lead ECG signals due to their simplicity and applicability in wearable devices. This paper presents a comprehensive review of hybrid approaches combining Fast Fourier Transform (FFT) and Continuous Wavelet Transform (CWT) with stochastic pooling-based neural network architectures to enhance detection accuracy. FFT captures global frequency-domain characteristics, while CWT provides localized time-frequency information, making their hybridization highly effective for AF feature extraction. Stochastic pooling improves generalization and prevents overfitting in deep networks. The review focuses on literature from 2020–2023, analyzing recent trends, methodologies, datasets, and performance metrics. Comparative analysis highlights the superiority of hybrid transform-based deep learning models over conventional machine learning approaches. Challenges such as noise sensitivity, data imbalance, and real-time deployment are discussed. The study concludes that integrating advanced signal processing techniques with optimized neural architectures significantly improves AF detection accuracy and robustness in modern healthcare systems.