A Deep Learning Method to Detect Atrial Fibrillation Based on Continuous Wavelet Transform.

Area of Science:

• Cardiology
• Artificial Intelligence
• Signal Processing

Background:

• Atrial fibrillation (AF) is a prevalent cardiac arrhythmia with significant clinical implications.
• Accurate and automated detection of AF remains a challenge in clinical practice and research.
• Existing methods for AF detection often require complex feature engineering or are limited in scope.

Purpose of the Study:

• To propose and evaluate a novel deep learning-based method for the automatic detection of atrial fibrillation (AF) from single-lead electrocardiogram (ECG) signals.
• To assess the efficacy of the proposed method across different rhythm classifications, including normal sinus rhythm (NSR), AF, other arrhythmias (OTHER), and noise (NOISE).
• To investigate the performance of the deep learning model using various wavelet bases for signal transformation.

Main Methods:

• A deep learning approach utilizing a convolutional neural network (CNN) was developed for AF detection.
• Continuous Wavelet Transform (CWT) was employed to convert 10-second single-lead ECG signals into wavelet coefficient matrices.
• The CNN model was trained on a multi-database ECG dataset categorized into NSR, AF, OTHER, and NOISE classes.

Main Results:

• The proposed deep learning method demonstrated high performance in detecting AF.
• When using a Morlet wavelet, the method achieved an overall accuracy of 97.56%.
• The model exhibited an average sensitivity of 97.56%, an average specificity of 99.19%, and an Area Under the Curve (AUC) of 0.9983.

Conclusions:

• The developed deep learning method, leveraging CWT and CNN, is effective for the automatic detection of atrial fibrillation from ECG signals.
• The use of a Morlet wavelet base with the proposed architecture yields promising and highly accurate results for AF classification.
• This novel approach holds potential for improving the clinical diagnosis and management of patients with AF.