Automatic detection of atrial fibrillation using stationary wavelet transform and support vector machine.

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Atrial fibrillation (AF) is a prevalent cardiac arrhythmia causing significant public health issues.
• Early detection of AF is crucial for managing complications and improving patient prognosis.
• Existing AF detection methods often rely on beat detection, which can impact overall performance.

Purpose of the Study:

• To propose a novel method for automatic detection of atrial fibrillation (AF).
• To develop an AF detection system that does not depend on P-peak or R-Peak detection.
• To evaluate the performance of the proposed method against existing algorithms.

Main Methods:

• Employed stationary wavelet transform and support vector machine for AF episode detection.
• Eliminated the need for P-peak or R-Peak detection, simplifying the pre-processing step.
• Compared the proposed method with existing techniques using the MIT-BIH Atrial Fibrillation database.

Main Results:

• Achieved an area under the ROC curve of 99.5% via stratified 2-fold cross-validation.
• Demonstrated high accuracy (sensitivity 97.0%, specificity 97.1%) even with short data segments (10s).
• Maintained high performance irrespective of parameter value choices.

Conclusions:

• The proposed AF detection method exhibits high sensitivity and specificity.
• The method's robustness and accuracy make it suitable for practical clinical applications.
• This novel approach offers a promising tool for automated atrial fibrillation detection.